wenmin-wu
535 verified skills14,723 total stars
cv-lap-hard-negative-mining
Use linear assignment problem (LAP/lapjv) on a score matrix to select globally optimal hard-negative pairs for metric learning
data-ai31
llm-sliding-window-permutation-search
Local search that slides a window of size p across a word sequence, brute-forcing all permutations within each window to minimize an objective like LLM perplexity
testing31
llm-perplexity-prompt-ranking
Rank candidate prompts by computing LLM perplexity of the full conversation conditioned on each prompt, selecting the lowest-perplexity candidate as the best match
data-ai31
llm-greedy-word-reinsert-search
Greedy local search that removes one element from a fixed position and re-inserts it at every possible index, keeping the best improvement per round
data-ai31
llm-batched-perplexity-scoring
Batch-compute perplexity for multiple texts using a causal LM with proper padding, shifted labels, and pad-token masking for efficient GPU utilization
testing31
cv-slice-padding-augmented-duplicates
Pad 3D volumes with fewer slices than required by duplicating existing slices with slight brightness variation via convertScaleAbs
tools31
cv-batch-all-contrastive-loss
All-vs-all contrastive loss comparing every pair in a batch (N^2 pairs) with margin and compactification regularizer
data-ai31
nlp-anchor-grouped-validation
Split validation by unique anchor/query entities so no anchor appears in both train and val, preventing data leakage in pairwise matching tasks
data-ai31
tabular-distribution-matching-postprocess
Reshapes model predictions to match the known label distribution from training data using rank-based mapping.
data-ai31
tabular-frequency-encoding
Adds each feature's value-count frequency as a new column, enabling tree models to split on how common or rare a value is.
data-ai31
tabular-group-kfold-leak-prevention
Uses GroupKFold to prevent data leakage when multiple rows share a common entity (e.g., same user, question, or document).
documentation31
tabular-group-shuffle-split
Splits train/validation using GroupShuffleSplit so that related samples (forks, families, sessions) never span both sets.
data-ai31
tabular-lgbm-dart-boosting
Use LightGBM DART boosting (dropout on trees) with aggressive feature and bagging fractions to reduce overfitting on high-dimensional tabular data
data-ai31
tabular-log-odds-fold-averaging
Average model predictions across CV folds in log-odds space rather than probability space for better-calibrated ensemble outputs
data-ai31
tabular-null-importance-feature-selection
Scores features by comparing actual importances against a null distribution from shuffled targets, removing features that cannot beat random noise.
data-ai31
tabular-pairwise-te-logit-stacking
Generates all C(n,2) pairwise feature combinations, target-encodes each pair with cuML TargetEncoder, then applies logit polynomial expansion (z, z^2, z^3) for stacking with cuML LogisticRegression.
development31
tabular-pseudo-labeling
Augments training data with high-confidence test predictions as pseudo labels, retrains the model, and keeps the result only if OOF AUC improves. A semi-supervised technique for tabular competitions.
testing31
tabular-polynomial-interaction-features
Generates polynomial powers and interaction terms from selected numeric features to capture nonlinear relationships with the target.
tools31
tabular-row-aggregate-features
Engineers row-wise statistical features (sum, mean, std, skew, kurtosis, median, min, max) across all numeric columns per sample.
content-media31
tabular-row-wise-target-normalization
Normalizes each sample's multi-output target vector to zero mean and unit variance, removing per-sample scale differences before training.
data-ai31
tabular-spatial-distance-aggregation
Compute min/max/mean/std of Euclidean distances from all entities to a key point, then aggregate per group for spatial feature engineering
tools31
tabular-synthetic-sample-detection
Detects synthetic/fake test samples by checking whether each row has at least one unique value across all features — real samples do, synthetic ones don't.
testing31
timeseries-prediction-smoothing-lpf
Applies rolling mean or Butterworth low-pass filter to model predictions for temporal consistency and noise reduction.
data-ai31
tabular-weather-ordinal-encoding
Map free-text categorical descriptions to ordinal numeric scores via keyword matching — captures ordered severity in a single dense feature
development31
tabular-weighted-recall-multi-objective-metric
Evaluate recommendation quality with recall@K per action type, combined via business-importance weights
testing31
timeseries-cnn-transformer-multimodal-fusion
Process multiple sensor modalities through separate CNN branches then fuse via a transformer with CLS token for classification
tools31
timeseries-dilated-conv-residual-gru
Combines dilated 1D convolutions for multi-scale receptive fields with residual bidirectional GRU layers for sequence classification.
tools31
timeseries-event-peak-detection
Detects discrete events (state transitions) from continuous predictions using local maxima with minimum-interval constraints.
data-ai31
timeseries-k-mode-gaussian-nll-loss
Negative log-likelihood loss over K isotropic-Gaussian trajectory modes with per-mode confidences and logsumexp stability
tools31
timeseries-mean-residual-decomposition
Decompose multi-output prediction into a global mean (1D model) plus per-channel residuals (2D model) with quadrature uncertainty
data-ai31
timeseries-mixup-sequence-augmentation
Apply MixUp augmentation to padded time-series batches with Beta-distributed lambda and soft label mixing
tools31
timeseries-se-residual-1d-cnn
1D ResNet block with Squeeze-and-Excitation channel attention for temporal sensor feature extraction
testing31
tabular-dtype-preset-csv-load
Predefines minimal unsigned integer dtypes before CSV loading to cut DataFrame memory usage by 2-4x without any data loss.
data-ai31
timeseries-scaled-pinball-loss
Scaled Pinball Loss (SPL) metric for evaluating quantile forecasts, normalized by mean absolute successive differences of training data
data-ai31
timeseries-ratio-target-for-smape
Transform forecasting target to next/current ratio minus one so that optimizing MAE or squared error implicitly minimizes SMAPE
testing31
timeseries-retroactive-outlier-rescaling
Walk backward through a time series and multiplicatively rescale segments when jumps exceed a fraction of the running mean to correct data collection anomalies
data-ai31
tabular-simulated-annealing-multi-operator
Simulated annealing with diverse move operators (translate, rotate, swap, Levy flight, squeeze) and adaptive reheating on stagnation for combinatorial optimization
tools31
timeseries-density-to-count-roundtrip
Convert a density metric back to integer counts using known population, round to nearest integer, then recompute density to exploit the discrete nature of the target
tools31
tabular-rmsle-keras-metric
Custom Keras RMSLE metric using K.log with K.clip to safely evaluate price and count regression during training
tools31
cv-cnn-rnn-video-classification
Extract per-frame CNN features then classify the temporal sequence with stacked GRU layers and a boolean mask for variable-length video inputs
content-media31
cv-deterministic-hash-partitioning
Partition a large dataset into N balanced shards using integer key modulo arithmetic for reproducible, class-interleaved splits across CSV files
testing31
cv-clip-interrogator-captioning
Generate descriptive text prompts from images by combining BLIP captioning with CLIP cosine similarity against curated label banks for medium, movement, and flavor attributes
tools31
cv-embedding-knn-regression
GPU-accelerated k-NN regression on CLIP image embeddings using cosine distance and inverse-distance-power weighting to predict target embedding vectors
tools31
tabular-opponent-avoidance-scoring
Score candidate movement directions by average distance to nearby opponents and pick the safest path for ball-carrying agents in game AI
data-ai31
tabular-featureunion-field-dispatch
Use sklearn FeatureUnion with closure-based preprocessors to apply different vectorizers to different DataFrame columns in a single fit_transform call
data-ai31
tabular-brand-name-recovery-from-title
Recover missing categorical values by matching words in a related text field against a known vocabulary built from the full dataset
development31
tabular-ball-landing-prediction
Predict where a projectile will land using kinematic equations with estimated gravity to intercept aerial passes in game AI simulations
data-ai31
nlp-lda-topic-modeling
Latent Dirichlet Allocation on CountVectorizer bag-of-words to discover latent topics with per-document topic distributions for feature engineering or EDA
documentation31
cv-sentence-transformer-target-encoding
Encode text prompts into fixed-length dense vectors using SentenceTransformer for cosine-similarity evaluation in image-to-text retrieval tasks
development31
cv-pairwise-tracking-feature-merge
Double left-join on tracking data to create pairwise features (positions, velocities, distance) for both entities in an interaction pair
testing31
cv-percentile-contrast-stretch
Normalize high-dynamic-range satellite or medical imagery to [0,1] using per-channel percentile clipping to suppress outliers while preserving relative contrast
tools31
cv-mask-to-polygon-contour-hierarchy
Convert binary segmentation masks to Shapely MultiPolygons using cv2 contour hierarchy to correctly handle interior holes, with Douglas-Peucker simplification
development31
cv-3d-center-crop-zero-pad
Select center slices from a 3D volume and zero-pad along depth when the scan has fewer slices than the target count
tools31
timeseries-tof-spatial-region-pooling
Aggregate high-dimensional spatial sensor grids into hierarchical region statistics at multiple granularities
tools31
timeseries-sensor-modality-dropout
Randomly zero out entire sensor modalities during training with a learned gate to handle missing modalities at inference
data-ai31
timeseries-poisson-lgbm-bias-corrected-ensemble
Train a single Poisson LightGBM count forecaster, then ensemble its predictions with multiple multiplicative scaling factors (alpha ≈ 1.02-1.03) to undo the systematic downward bias of Poisson regression on intermittent retail data
data-ai31
timeseries-quaternion-angular-velocity
Derive angular velocity from consecutive quaternion frames via relative rotation and rotvec conversion
testing31
timeseries-neighbor-average-nan-interpolation
Fill gaps in a daily exogenous series (oil prices, sensor feeds) by merging against a full calendar to expose NaNs, then replacing each NaN with the midpoint of its nearest valid left and right neighbors, walking outward past consecutive NaN runs
testing31
timeseries-multi-scale-rolling-features
Computes rolling mean/max/std at multiple window sizes plus total variation (abs first differences) for multi-resolution temporal context.
tools31
timeseries-multi-gap-lag-diff-features
Generate shift/diff features at multiple lag sizes (1,2,3,5,10,20,50,100) over cursor/time/state series, then aggregate statistics per session
tools31
timeseries-locale-scoped-holiday-flag
Build a single per-row "day off" boolean from a holidays table with National/Regional/Local locale hierarchy and Work Day overrides that flip make-up working weekends back to working days
development31
timeseries-inverse-variance-channel-weighting
Weight multi-channel signals by inverse per-channel variance with percentile clipping, emphasizing low-noise channels in aggregation
tools31
timeseries-gradient-transit-phase-detection
Detect event ingress/egress boundaries by finding steepest gradient on each side of the signal minimum in a smoothed time series
tools31
timeseries-gradient-event-boundary-detection
Detect event start/end boundaries in time series by finding extrema of the first derivative (steepest gradient points)
tools31
timeseries-event-anchored-frame-sync
Align low-Hz sensor data to high-fps video by anchoring a named event (e.g. ball_snap) to a known frame index and converting time offsets via fps
data-ai31
timeseries-detector-calibration-pipeline
Multi-step detector calibration pipeline — ADC inversion, hot/dead pixel masking, nonlinearity correction, dark subtraction, flat-field normalization
devops31
timeseries-correlated-double-sampling
Subtract paired reference frames from signal frames to cancel readout noise and common-mode bias
data-ai31
tabular-yearly-partitioned-groupby
Split a multi-year table into per-year partitions, run the same groupby aggregation on each, then concat and gc — a pure-pandas map-reduce that survives 100M+ rows on a 16GB kernel
tools31
tabular-xgb-gpu-batch-iterator
Use XGBoost DeviceQuantileDMatrix with a custom batch iterator to train on large datasets without exhausting GPU memory
development31
tabular-vote-ensemble-outer-join
Ensemble ranked recommendation lists by outer-joining exploded candidates and re-ranking by weighted vote sum
tools31
tabular-typed-panel-aggregation
Aggregate panel/sequential data with type-appropriate statistics — numeric (mean/std/min/max/last) and categorical (count/last/nunique) — then concat into flat features
development31
tabular-weighted-gini-top-recall-metric
Custom ranking metric combining normalized weighted Gini coefficient with top-K% capture rate for imbalanced classification with class-weighted evaluation
testing31
tabular-two-level-hierarchical-aggregation
Aggregates deeply nested relational tables through two groupby levels (child → intermediate → parent) to build features from multi-hop relationships.
development31
tabular-temporal-session-aggregation
Builds user-level features by accumulating statistics across sequential event sessions before each assessment point.
development31
tabular-tabular-to-image-cnn
Reshapes tabular features into 2D pseudo-images via random feature permutation, enabling CNN-based feature interaction learning.
content-media31
tabular-svd-target-reconstruction
Compresses high-dimensional targets with TruncatedSVD, trains on the reduced space, then reconstructs full predictions via the components matrix.
data-ai31
tabular-sparse-dense-hstack-lgbm
Train LightGBM directly on a scipy.sparse.hstack of TF-IDF text vectors and dense tabular columns, passing feature_name and categorical_feature so native categorical handling survives the sparse block
development31
tabular-regression-to-cdf-smoothing
Convert a scalar regression prediction into a smoothed CDF over discrete bins using a linear ramp instead of a hard step
tools31
tabular-rdkit-molecular-descriptors
Computes all numeric RDKit molecular descriptors from SMILES strings, filtering out NaN, constant, and infinite values to produce a clean feature matrix.
tools31
tabular-play-direction-normalization
Mirror spatial coordinates and angles so all plays face the same direction — removes left/right asymmetry from sports and spatial data
data-ai31
cv-black-slice-replacement
Detect all-black DICOM/MRI slices (mean==0) and replace them by randomly sampling a non-black slice from the same series
data-ai31
cv-brovey-pansharpening
Fuse low-resolution multispectral bands with a high-resolution panchromatic band using the Brovey transform to produce sharp multi-band imagery
content-media31
tabular-personnel-count-parsing
Parse structured text fields like '1 RB, 2 TE, 2 WR' into separate numeric columns per category
development31
cv-centroid-distance-anomaly-score
Score each face in a video as anomalous by computing L2 distance from the embedding centroid of all faces, then convert to probability via logistic function
data-ai31
cv-smm-spatial-observation-encoding
Encode game state as a Super Mini Map (SMM) with separate binary channels for players, ball, and ownership, bit-packed for efficient transfer in RL training
development31
tabular-per-feature-bias-correction
Post-processing correction for multi-output regression — scale each output by its train-derived mean ratio to fix systematic per-feature bias
data-ai31
tabular-per-partition-variance-filtering
Apply VarianceThreshold within each data partition on combined train+test to select informative features per subgroup
development31
cv-structured-gt-serialization-tokens
Serialize structured ground truth (chart data, table rows, form fields) as special-token-delimited sequences for generative vision-language model training
testing31
cv-tensor-core-aligned-padding
Pad batch sequence lengths to multiples of 8 for efficient tensor core utilization on GPUs, with -100 masking for label padding
testing31
cv-temporal-frame-jitter-augmentation
Add random temporal offset to the center frame during training to augment temporal diversity in video-based models
data-ai31
tabular-outlier-aware-two-stage-blending
When a regression target has a long discrete tail (e.g. ~1% of rows pinned at -33.22 in Elo), train one regressor on the *non-outlier* subset, a separate binary classifier for the outlier flag, and splice the predictions — replace the top-K most-confident outlier predictions in the regressor's output with the outlier value, where K is calibrated on validation
data-ai31
tabular-next-click-time-delta
Computes seconds until the next event within a group using diff().shift(-1) on sorted timestamps, capturing user behavior velocity.
tools31
llm-sentence-truncation-fallback
Truncate LLM output to exactly N sentences and fall back to a known-good baseline string when output is empty or too short
development31
llm-spiral-patrol-exploration
Agents patrol in expanding spiral patterns using rotating direction sequences with increasing radius for systematic grid exploration
data-ai31
tabular-hierarchical-collision-cascade
Three-level polygon overlap test — AABB early exit, then point-in-polygon ray casting, then segment intersection — for fast non-convex collision detection
development31
tabular-multi-source-candidate-fusion
Fuse recommendation candidates from user history, multiple co-visitation matrices, and global popularity in a priority-ordered cascade
tools31
tabular-multi-input-embedding-nn
Keras multi-input model with separate embedding layers for categoricals, GRU for text sequences, and dense layers for numerics, all concatenated into a shared regression trunk
development31
tabular-leak-free-loop-features
Iterates through rows chronologically to accumulate user statistics, fetching current state before updating to prevent future data leakage.
data-ai31
timeseries-kaggle-api-streaming-inference
Predict day-by-day via Kaggle's iter_test API while maintaining a rolling history buffer for computing lag features online
development31
timeseries-multi-lag-target-features
Generate lag features for multiple targets over N days by shifting evaluation dates and self-joining per entity, creating a wide feature matrix of past target values
tools31
tabular-implicit-als-collaborative-filtering
Alternating Least Squares matrix factorization on sparse user-item interaction matrices for implicit feedback recommendations.
testing31
tabular-cyclical-feature-encoding
Encodes cyclical features (hour, month, day-of-week) using sine/cosine transforms to preserve circular distance.
tools31
nlp-pearson-correlation-metric
Use Pearson correlation coefficient as evaluation metric for semantic similarity regression tasks, selecting best checkpoint by correlation rather than loss
testing31
nlp-two-stage-retrieve-rerank
Two-stage pipeline where an unsupervised bi-encoder retrieves KNN candidates and a supervised cross-encoder reranks them with sigmoid thresholding
development31
nlp-taxonomy-context-enrichment
Enrich model input by mapping categorical codes to human-readable taxonomy descriptions and concatenating them as context for transformer models
development31
nlp-dropout-disabled-inference
Explicitly zero all dropout probabilities in transformer config at load time for fully deterministic inference
testing31
cv-learned-distance-metric
Trainable nonlinear distance metric that transforms (v1-v2) and (v1-v2)^2 through a linear layer before computing squared norm
data-ai31
cv-greedy-mask-overlap-resolution
Resolve overlapping instance masks by greedily assigning contested pixels to higher-confidence predictions using a running occupancy map
testing31
cv-cumulative-sum-channel
Add a cumulative sum channel along the vertical axis to capture directional structural trends in grayscale images for segmentation
content-media31
tabular-threaded-parquet-describe-features
Parallel-load per-subject parquet time-series files with ThreadPoolExecutor and flatten describe() statistics into tabular feature vectors
data-ai31
tabular-time-varying-reward-shaping
Shape RL rewards with time-decaying asset weights and time-increasing resource weights so the agent transitions from expansion to accumulation as the game progresses
development31
tabular-toroidal-manhattan-distance
Compute shortest Manhattan distance on a toroidal (wrapping) grid by comparing normal vs wrap-around routes in each axis
testing31
tabular-transductive-train-test-transform
Fit unsupervised transforms (scaler, PCA, variance filter) on combined train+test data for more stable statistics, especially on small datasets
testing31
tabular-transitive-match-closure
Post-processes entity match predictions to enforce symmetry (A→B implies B→A) and transitivity (A→B, B→C implies A→C) via graph closure.
tools31
cv-tumor-volume-ratio-features
Extract volumetric features from 3D segmentation masks including scan/tumor pixel ratios, tumor percentage, and tumor centroid coordinates
tools31
llm-turn-based-prompt-accumulation
Rebuild multi-turn conversation context by interleaving user/assistant turns with chat template tokens into a single prompt each call
development31
timeseries-tweedie-objective-zero-inflated
Use LightGBM's tweedie objective with variance_power between 1.05 and 1.2 for zero-inflated count forecasting (retail SKUs, intermittent demand, click events) — handles the "many zeros plus a heavy right tail" distribution that breaks both regression (RMSE) and classification (BCE) objectives
tools31
cv-two-stage-bce-then-lovasz
Train segmentation with BCE loss first for stable convergence, then fine-tune with Lovasz-hinge on raw logits for IoU-optimal predictions
data-ai31
timeseries-unknown-class-residual-probability
Estimate out-of-distribution class probability as the product of (1 - p_i) across all known classes, scaled by a calibrated prior
tools31
cv-video-frame-sampling-pipeline
Efficiently sample N evenly-spaced frames from a video using OpenCV grab/retrieve pattern with optional resize for batch face detection or classification
testing31
timeseries-wavelet-denoising
Denoise an erratic 1D series with discrete wavelet decomposition + universal soft thresholding (sigma estimated from MAD of the detail coefficients) to extract the underlying trend/seasonality without lagging the signal — a far better trend extractor than rolling means for spiky retail or sensor data
testing31
cv-weighted-embedding-blending
Ensemble predictions from heterogeneous vision-language models by blending their output embeddings with fixed scalar weights in embedding space
data-ai31
tabular-weighted-position-decay-ensemble
Ensembles multiple ranked recommendation lists by scoring items as model_weight / position_rank, then re-ranking.
data-ai31
cv-coco-annotation-conversion
Convert per-instance RLE or polygon annotations to COCO JSON format for seamless use with Detectron2 and MMDetection
development31
llm-code-execution-weighted-vote
Execute Python code blocks from LLM math responses in a sandbox, then double-weight code-derived answers in majority voting
development31
tabular-collinear-feature-removal
Removes redundant features by iterating pairwise Pearson correlations and dropping one member of each pair exceeding a threshold.
data-ai31
tabular-game-state-grid-encoding
Encode a 2D game board into a normalized multi-channel feature tensor with log-scaled resources, signed unit counts, and directional features for RL agents
development31
cv-per-modality-ensemble-averaging
Train separate models per imaging modality (FLAIR/T1w/T1wCE/T2w) and average their predictions for final ensemble
data-ai31
tabular-iterative-pseudo-label-refinement
Multi-round pseudo labeling with progressively confident test predictions merged into training plus OOF-based train label correction
testing31
cv-3d-encoder-2d-decoder-segmentation
3D ResNet encoder extracts volumetric features, pools depth dimension, then feeds into a 2D UNet/FPN decoder for segmentation
development31
llm-4bit-nf4-double-quantization
Load large LLMs with 4-bit NF4 quantization and optional double quantization via BitsAndBytes to reduce GPU memory by 4x while preserving inference quality
testing31
timeseries-activity-threshold-lastval-fallback
Override model predictions with last known value for low-activity or low-density entities where learned trends are unreliable
data-ai31
llm-actor-critic-game-agent
Shared-backbone neural network with actor (policy) and critic (value) heads for grid-based game agent RL training
data-ai31
llm-adaptive-time-budget-inference
Dynamically reduce max_tokens and batch size as wall-clock time approaches a cutoff to ensure all inputs get processed
data-ai31
llm-answer-accumulation-filter
Parse sequential yes/no answers to build inclusion/exclusion sets, then apply compound boolean filters to narrow a candidate list
development31
tabular-autoencoder-timeseries-embedding
Train a PyTorch autoencoder on time-series summary statistics to produce dense encoded features for downstream GBDT models
development31
timeseries-availability-masked-regression-loss
Multiply per-timestep regression loss by a 0/1 availability mask so missing future steps contribute zero gradient
data-ai31
cv-bbox-interpolation-temporal-tracking
Interpolate missing bounding boxes across video frames using bidirectional pandas interpolation to maintain smooth tracking through occlusions
data-ai31
nlp-best-prob-fallback-matching
When no candidate passes the threshold for a query, fall back to the single highest-scoring match to guarantee at least one prediction per query
databases31
llm-binary-answer-clamping
Force-clamp free-form LLM output to binary yes/no with keyword matching and a fallback default for constrained environments
data-ai31
llm-binary-search-entity-narrowing
Hierarchical binary search over entity space by asking category, region, then first-letter questions to narrow candidates before guessing
development31
timeseries-bootstrapped-residual-prediction-intervals
Generate prediction intervals by repeatedly sampling from model residuals, adding to point forecasts, and taking quantiles across synthetic futures
data-ai31
llm-boxed-answer-extraction
Extract final numeric answers from LaTeX \boxed{} notation in LLM math reasoning output, scanning matches in reverse for robustness
data-ai31
timeseries-burst-rle-detection
Detect P-bursts (fast-typing runs) and R-bursts (consecutive revisions) via polars run-length encoding over boolean event conditions
tools31
llm-cargo-threshold-state-machine
Per-agent COLLECT/DEPOSIT state machine driven by cargo thresholds with greedy neighbor selection for resource collection games
development31
cv-center-z-slice-selection
Select a fixed number of Z-slices centered around the volume midpoint for memory-efficient 2.5D input from 3D CT/MRI stacks
testing31
tabular-chained-target-prediction
Predict correlated targets sequentially, using earlier target predictions as input features for subsequent targets to exploit inter-target dependencies
testing31
cv-chunked-csv-image-generator
Memory-efficient Keras generator that streams sharded CSV files in chunks, renders strokes to images on-the-fly, and yields batches for training on datasets too large for memory
testing31
cv-class-density-threshold-patch-sampling
Sample training patches from large images using per-class area-fraction thresholds to ensure each patch contains meaningful object coverage
data-ai31
cv-classifier-free-guidance-diffusion
Manual stable diffusion inference loop with classifier-free guidance that interpolates between unconditional and conditional noise predictions for controllable image generation
content-media31
timeseries-class-weighted-multiclass-logloss
Custom multiclass log-loss that weights per-class contributions by class frequency and domain importance, usable as both training loss and eval metric
data-ai31
tabular-column-shuffle-augmentation
Augments imbalanced tabular data by independently shuffling each feature column within a class, creating synthetic samples that preserve per-column marginal distributions.
data-ai31
tabular-confidence-probability-clipping
Hard-clips predicted probabilities to 0 or 1 when they exceed high-confidence thresholds, reducing log loss on near-certain predictions.
tools31
tabular-confidence-weighted-rate-encoding
Encodes categorical groups by their target rate scaled by a log-confidence factor, smoothing unreliable rates from low-frequency groups toward zero.
development31
tabular-content-difficulty-features
Precomputes item/content difficulty as historical mean accuracy, merged as a static feature for user-item prediction tasks.
content-media31
cv-conv1d-lstm-stroke-classifier
Stack 1D convolutions for local feature extraction before bidirectional LSTMs to classify variable-length stroke sequences into hundreds of doodle categories
development31
tabular-co-purchase-item-pairing
Recommends items frequently purchased together with a customer's recent items using pre-computed pair dictionaries.
data-ai31
cv-coverage-stratified-split
Stratify train/validation split by binned mask coverage percentage to ensure balanced foreground representation in segmentation tasks
data-ai31
tabular-cross-dataset-user-aggregation
Build user-level behavioral features (avg listing duration, relisting frequency, total items) by joining auxiliary activity tables that share user_id but not item_id with train/test
development31
tabular-crps-cdf-loss
Model cumulative distribution via softmax output layer and CRPS loss — for probabilistic regression over discrete bins
data-ai31
cv-isotropic-resize-with-padding
Resize images preserving aspect ratio then zero-pad to a square to avoid distortion artifacts in face crops or object detection inputs
content-media31
cv-detectron2-custom-data-mapper
Custom Detectron2 data mapper with photometric augmentations that properly transforms images, bounding boxes, and instance masks in sync
data-ai31
llm-inverse-task-prompt-template
Structured prompt template for recovering the instruction that transformed one text into another, with labeled original/rewritten fields and explicit task framing
testing31
tabular-declarative-groupby-aggregation
Config-driven feature factory that generates groupby aggregation features from a declarative spec list, supporting count, mean, var, nunique, cumcount, and custom lambdas.
tools31
nlp-domain-special-token-embedding
Add domain-specific categorical values as new special tokens, resize embeddings, and prepend them to input so the model learns domain-aware representations
development31
llm-dual-role-agent-dispatch
Single agent entry point that dispatches between multiple roles (ask/answer/guess) based on turn type, combining heuristic and LLM-based strategies
data-ai31
cv-dual-view-reshape-forward
Reshape dual-view stacked channels into doubled batch dimension for shared backbone, then concatenate with tabular features for classification
tools31
timeseries-mc-dropout-uncertainty
Estimate prediction uncertainty via Monte Carlo Dropout — run inference N times with dropout active and compute mean/std
data-ai31
llm-exhaustive-permutation-early-stopping
Enumerate all factorial permutations in batches with LLM scoring, tracking the running best and early-stopping when score crosses a known optimality threshold
data-ai31
timeseries-expanding-window-stacking
Walk-forward stacking ensemble that trains base models on expanding windows and a meta-learner on their out-of-fold predictions across time
data-ai31
timeseries-flux-snr-weighted-features
Engineer SNR-derived features from irregular time series — flux ratio squared, error-weighted mean flux, and normalized amplitude/range features
tools31
cv-frame-differencing-temporal-encoding
Encode motion and velocity by computing per-channel pixel differences between consecutive frames instead of stacking raw frames for RL visual observations
development31
timeseries-gaussian-log-likelihood-metric
Evaluate probabilistic forecasts using normalized Gaussian log-likelihood relative to naive and oracle baselines, scoring both mean accuracy and uncertainty calibration
data-ai31
cv-generative-output-numeric-cleaning
Clean noisy numeric strings from generative model output by removing invalid characters, fixing malformed floats, and handling multiple decimal points
testing31
tabular-gmm-feature-augmentation
Fits a Gaussian Mixture Model on the joint feature-target space and samples synthetic data pairs to augment small tabular datasets.
data-ai31
tabular-group-mean-log-mae-metric
Custom evaluation metric that computes log of per-group MAE then averages, penalizing uniformly bad groups.
data-ai31
tabular-gnn-on-knn-graph
Constructs a customer similarity graph via KNN on mixed features, then trains a GraphSAGE GNN for node classification. Captures relational patterns that tree and linear models miss, adding ensemble diversity.
data-ai31
tabular-haversine-knn-candidate-generation
Generates geographically proximate candidate pairs for entity matching using KNN with haversine distance, optionally partitioned by country.
data-ai31
tabular-hierarchical-rule-engine
Two-level group-then-pattern dispatch for game AI agents where groups filter by game state and ordered patterns within a group fire the first matching action
data-ai31
timeseries-hierarchy-level-confidence-coefficients
Assign different uncertainty spread coefficients per aggregation level in hierarchical forecasts, reflecting that higher aggregation yields narrower intervals
testing31
timeseries-imu-gravity-removal
Remove gravity component from raw accelerometer data using quaternion rotation to yield linear acceleration
data-ai31
tabular-inner-kfold-target-encoding
Computes leak-free target encoding statistics (mean, std, min, max) using nested inner KFold within each outer CV fold, preventing target leakage that occurs with naive groupby-based encoding.
data-ai31
llm-low-meaning-input-substitution
Replace the actual input text with a generic low-meaning passage to prevent the LLM from fixating on content specifics, forcing it to focus on stylistic and structural transformation cues
development31
timeseries-keras-multi-quantile-loss
Single neural network outputting all quantiles simultaneously via pinball loss over a quantile vector for joint probabilistic forecasting
data-ai31
timeseries-keystroke-pause-bucket-features
Bucket inter-keystroke latencies into pause-duration ranges (0.5-1s, 1-1.5s, 1.5-2s, 2-3s, >3s) and count per session as hesitation features
testing31
timeseries-learnable-fir-filter
Initialize a depthwise Conv1d with FIR filter coefficients as a trainable high-pass/low-pass filter for sensor signal preprocessing
testing31
tabular-multi-seed-fold-averaging
Trains multiple models per CV fold with different random seeds for augmentation, then averages their predictions to reduce variance from stochastic data generation.
data-ai31
tabular-l1-coefficient-interaction-map
Extract and visualize per-subgroup feature coefficient signs from L1-regularized models as an interaction heatmap for EDA
testing31
tabular-last-diff-lag-features
Compute first-order difference between last and second-to-last rows per entity in panel data to capture recent trend direction and magnitude
data-ai31
nlp-learned-attention-pooling
Replace mean pooling with a trainable attention network (Linear-Tanh-Linear-Softmax) that learns token importance weights over transformer hidden states
data-ai31
nlp-length-sorted-batching
Sort texts by length before batching with dynamic padding to minimize wasted padding tokens and speed up transformer inference
tools31
tabular-logit-transform-stacking
Applies logit transformation to base model probabilities before fitting a logistic regression meta-learner, enabling principled linear combination in log-odds space.
development31
tabular-logspace-recency-reranking
Rerank session candidates using log-spaced recency weights multiplied by interaction-type multipliers
tools31
tabular-majority-vote-submission-blend
Blend multiple submission CSVs by row-wise majority voting on discrete predictions to produce a more robust final output
tools31
cv-map-iou-precision-sweep
Compute mean Average Precision by sweeping IoU thresholds from 0.5 to 0.95 on RLE-encoded instance masks using pycocotools
tools31
tabular-multiclass-to-binary-collapse
Trains on a finer-grained multiclass target (subtypes), then collapses non-baseline classes into a single positive class for binary submission.
data-ai31
cv-metadata-injection-bottleneck
Inject scalar metadata (depth, position, clinical features) into U-Net bottleneck via RepeatVector and Reshape for metadata-aware segmentation
tools31
tabular-morgan-fingerprint-features
Converts molecular SMILES strings to fixed-length Morgan fingerprint bit vectors using RDKit for use as tabular ML features.
data-ai31
timeseries-multiband-color-index-features
Compute log-ratio features between adjacent frequency bands as color indices to characterize spectral shape from multi-band time series
tools31
timeseries-multiband-lomb-scargle-period
Estimate periodicity from irregularly sampled multi-band time series using the multiband Lomb-Scargle periodogram, then phase-fold observations
testing31
timeseries-multiband-tsfresh-fft
Apply tsfresh per-passband feature extraction with FFT coefficients to capture multi-band periodicity from irregular time series
testing31
timeseries-multimodal-trajectory-head
Single linear head that jointly predicts K candidate trajectories and K softmax confidences, sliced and reshaped for multimodal regression
tools31
cv-spatial-rect-train-val-split
Split large-image segmentation data into train/val by spatial rectangle regions with border buffer exclusion to prevent patch leakage
data-ai31
tabular-multi-output-auxiliary-targets
Neural network with multiple output heads for main target plus auxiliary targets, improving representation learning via shared layers.
data-ai31
nlp-multi-retriever-union-ensemble
Run multiple independent retrieve-rerank pipelines and union-merge their predicted IDs per query via explode-groupby-unique
devops31
tabular-nelder-mead-threshold-optimization
Use scipy Nelder-Mead simplex to optimize regression-to-ordinal thresholds maximizing quadratic weighted kappa on OOF predictions
testing31
tabular-popularity-fallback-recommendation
Fills unfilled recommendation slots with globally popular recent items to handle cold-start users and short lists.
tools31
tabular-ppo-gym-wrapper-kaggle-env
Wrap a Kaggle competitive game environment as an OpenAI Gym env with continuous action space for training PPO agents via stable-baselines3
data-ai31
tabular-predicted-class-mass-reweighting
Post-hoc rescales ensemble probabilities by the inverse of each class's estimated total mass across the test set, correcting for class imbalance in predictions.
testing31
tabular-ngram-composite-features
Creates bi-gram and tri-gram composite categorical features by concatenating top categorical columns, then target-encodes the composites. Captures interaction effects that tree models may miss.
development31
cv-numeric-categorical-auto-detection
Auto-detect whether a generated data series is numeric or categorical by measuring the fraction of digit characters in the concatenated values
development31
tabular-oof-meta-features
Generates out-of-fold predictions from auxiliary models and uses them as input features for the final model.
data-ai31
cv-open-set-distance-cutoff
Assign an unknown/novel class when all nearest-neighbor distances exceed a tuned cutoff threshold for open-set recognition
tools31
cv-per-class-score-threshold
Apply class-specific confidence thresholds by inferring the dominant class per image and indexing into a per-class threshold array
testing31
tabular-optuna-lgbm-tuning
Uses Optuna with TPE sampler for Bayesian hyperparameter optimization of LightGBM, searching key params like num_leaves, depth, and learning rate.
tools31
tabular-outlier-rate-label-encoding
Encode a categorical column by replacing each category with the per-category outlier rate (mean of a binary outlier flag), out-of-fold to avoid leakage — a target-aware encoding tuned to long-tail / sentinel-target problems where a binary classifier signal is more useful than the raw regression mean
development31
cv-pairwise-distance-proximity-filter
Compute Euclidean distance between entity pairs from tracking data and filter out pairs beyond a threshold to reduce inference candidates
tools31
tabular-pearson-correlation-loss
Uses negative row-wise Pearson correlation as a differentiable loss function for multi-output regression, directly optimizing the competition metric.
tools31
tabular-per-fold-threshold-voting-ensemble
Binarizes each CV fold's predictions using its own optimized threshold, then majority-votes across folds instead of averaging raw probabilities.
data-ai31
timeseries-periodogram-seasonality-detection
Use scipy periodogram to identify dominant seasonal frequencies in a time series before selecting Fourier feature orders or ARIMA seasonal parameters
testing31
tabular-per-target-nan-mask-training
Trains independent models per target by masking NaN labels, enabling multi-output regression on datasets where each target has different coverage.
data-ai31
tabular-per-type-model-training
Trains separate models for each discrete category (e.g., molecule type, product class) to capture type-specific patterns.
development31
tabular-phase-based-strategy-cycling
Divide a game into repeating phases (attack, mine, spawn) with turn-modular gating so the agent cycles between aggressive and economic behavior
tools31
tabular-type-weighted-covisitation-matrix
Build item co-visitation matrix from session pairs within a time window, weighting by interaction type (click/cart/order) via GPU self-join
tools31
cv-prediction-map-stitching-averaging
Stitch overlapping tile predictions into a full-resolution output by accumulating probabilities and dividing by per-pixel overlap counts
data-ai31
tabular-rectangular-flight-plan-encoding
Encode closed rectangular patrol routes as compact direction-distance strings for fleet pathfinding on toroidal game grids
development31
tabular-recursive-feature-elimination
Uses RFE with a tree estimator to iteratively remove least important features, selecting an optimal compact feature set.
tools31
tabular-prior-rebalancing-oversampling
Rebalances training data by oversampling the majority class to match a known test-set class prior, reducing prediction miscalibration.
testing31
cv-progressive-dropout-unet
Apply lower dropout in shallow/final U-Net layers and higher dropout in deep layers to preserve spatial detail while regularizing abstract features
data-ai31
llm-prompt-variant-ensemble
Generate multiple LLM responses using diverse system prompt variants to increase reasoning diversity for self-consistency voting
data-ai31
timeseries-quantile-ratio-scaling
Convert point forecasts to prediction intervals by scaling with logit-transformed quantile ratios passed through a Normal CDF
tools31
tabular-rank-averaging-ensemble
Ensembles multiple model predictions by converting to ranks, averaging, and normalizing back to [0,1].
data-ai31
tabular-rank-calibrated-blending
Blends predictions from multiple models by converting to ranks, weighting, and calibrating back to probabilities via rank-group means from a reference model. Ensures monotonic calibrated output.
data-ai31
tabular-recency-weighted-candidate-generation
Generates recommendation candidates by ranking a customer's purchase history by frequency and recency within a recent window.
tools31
timeseries-rolling-refit-arima-forecast
Walk-forward validation for ARIMA by refitting on history at each step, forecasting one step ahead, then appending the true observation — produces an honest one-step error distribution that mirrors nightly-retrained production forecasters
data-ai31
timeseries-recursive-multistep-forecasting
Forecast a multi-step horizon by predicting one day ahead, writing the prediction back into the panel as the new "actual", recomputing all lag and rolling features that depend on it, then predicting the next day — turns a one-step LightGBM regressor into a 28-day forecaster without changing the model
data-ai31
tabular-regression-to-ordinal-thresholding
Converts regression predictions to ordinal classes by optimizing bin thresholds to maximize Quadratic Weighted Kappa.
tools31
tabular-regularized-qda-classifier
Use QuadraticDiscriminantAnalysis with regularization for binary classification on data with Gaussian cluster structure
data-ai31
tabular-relative-deviation-features
Computes differences and ratios between group-level aggregates and raw values to capture how each sample deviates from its group.
development31
tabular-ridge-xgb-stacking
Two-stage stacking where Ridge regression on OHE+scaled features produces OOF predictions fed as an extra feature to XGBoost, letting the tree model correct non-linear residuals on top of captured linear patterns.
data-ai31
cv-siamese-pairwise-comparison-head
Siamese network head that compares two embeddings via element-wise multiply, add, abs-diff, and squared-diff features for verification tasks
data-ai31
timeseries-sigma-clip-outlier-masking
Detect and mask outlier data points using iterative sigma-clipping on reference frames or calibration data
tools31
cv-sigmoid-normalized-rmse
Sigmoid-transformed normalized RMSE that maps error from [0,inf) to a bounded (0,1] similarity score using R2-score ratio
tools31
cv-rotation-tta-segmentation
Test-time augmentation via 4 rotation angles (0/90/180/270), applying inverse rotation to each prediction before averaging
testing31
timeseries-scaled-nan-sentinel
Encode missing sensor data with a per-modality sentinel value that survives standardization and remains detectable after scaling
development31
timeseries-transit-depth-polynomial-optimization
Estimate event depth by optimizing a scalar scaling factor on the in-event segment that minimizes polynomial baseline residual across the full signal
tools31
tabular-season-phase-labeling
Map calendar dates to categorical season phases (offseason, preseason, regular, postseason) using np.select with boundary date conditions
development31
llm-self-consistency-majority-vote
Aggregate multiple LLM reasoning attempts via majority voting with random jitter tiebreaking and validity filtering
data-ai31
timeseries-sensor-phase-visualization
EDA visualization of multi-modal sensor data with axvspan shading for labeled behavioral phases and contiguous span detection
data-ai31
tabular-smiles-randomization-augmentation
Augments molecular datasets by generating multiple randomized SMILES strings for the same molecule, exploiting SMILES non-uniqueness to multiply training samples.
data-ai31
timeseries-snap-event-interaction-features
Build per-state SNAP / event-flag interaction features by multiplying the binary flag with the sales and revenue columns segmented by state, capturing the demand uplift on government-benefit days that affects only specific geographies and product categories
development31
tabular-tfidf-svd-dense-text-features
Compress TF-IDF sparse text vectors into a handful of dense TruncatedSVD components so GBDTs can consume free-text fields as plain tabular columns
data-ai31
tabular-tfidf-weighted-category-counts
Convert per-group categorical event counts into TF-IDF-style features using log(1+tf/total) * log(N/df)
development31
cv-spectral-index-segmentation
Compute normalized spectral band ratios (NDWI, CCCI, NDVI) from multispectral imagery and threshold for binary segmentation of water, vegetation, or other targets
testing31
tabular-squeeze-compact-local-search
Three-stage packing refinement — uniform squeeze toward centroid, greedy compaction per object, then multi-directional local search — to tighten solutions after metaheuristic optimization
tools31
timeseries-stateful-chunk-inference
Processes long sequences in fixed-size chunks while carrying RNN hidden state across chunks for memory-efficient inference.
testing31
llm-stopword-priority-sorting
Initialize text ordering by placing stopwords first then content words, producing low-perplexity starting points for combinatorial search over word permutations
testing31
timeseries-store-profile-hierarchical-clustering
Re-cluster retail stores by scale-normalized weekday/dayoff mean+std profiles using Ward agglomerative clustering, replacing vendor-supplied "type/cluster" labels that correlate with store size instead of demand shape
testing31
tabular-streaming-prediction-api
Online inference pattern that processes test batches sequentially, updating feature dictionaries incrementally for time-series prediction APIs.
development31
cv-stroke-normalize-simplify-pipeline
Normalize raw stroke coordinates to 0-255 range, resample at uniform arc-length spacing, then apply Ramer-Douglas-Peucker simplification
testing31
cv-stroke-temporal-color-rendering
Render stroke sequences to grayscale images with temporal intensity encoding where earlier strokes are brighter and later strokes fade to encode drawing order
development31
tabular-strtree-spatial-index-collision
Use Shapely STRtree spatial index for O(n log n) polygon overlap detection instead of brute-force O(n^2) pairwise checks
development31
llm-swarm-tactic-diversity
Assign each new agent a different directional rotation pattern from a set of permutations to ensure swarm coverage diversity across the map
data-ai31
tabular-successive-groupby-aggregates
Build hierarchical features for transaction panels by aggregating twice — first groupby (entity, sub-key) to get a per-(entity, sub-key) summary, then groupby (entity) on those summaries to compute mean/min/max/std across the sub-keys, capturing the *distribution* of per-customer behavior rather than a single flat mean
development31
tabular-tabnet-sklearn-wrapper
Wrap PyTorch TabNet in a scikit-learn BaseEstimator with built-in imputation and early stopping for use in VotingRegressor ensembles
testing31
tabular-tabpfn-small-dataset-ensemble
Ensembles TabPFN (a prior-fitted Bayesian transformer for small tabular data) with XGBoost, averaging probabilities for stronger predictions on datasets under 1000 rows.
data-ai31
timeseries-temporal-frame-binning
Reduce temporal resolution by averaging consecutive frame blocks to improve SNR and compress high-cadence data
data-ai31
tabular-convex-hull-bbox-rotation
Minimize axis-aligned bounding box side length by finding the optimal rotation angle over convex hull vertices using bounded scalar optimization
tools31
cv-frame-prediction-averaging
Average per-frame sigmoid predictions across sampled video frames to produce a stable video-level classification probability
data-ai31
cv-lateralized-label-flip-tta-disable
Disable horizontal-flip augmentation (both train-time and TTA) when label columns encode left/right anatomy — flipping silently corrupts the targets because "Left ICA" must map to "Right ICA" after a flip, not stay as "Left ICA"
development24
cv-levenshtein-distance-metric
Evaluates image-to-sequence models using mean Levenshtein edit distance between predicted and ground-truth strings.
data-ai24
cv-lovasz-hinge-loss
Lovasz hinge loss that directly optimizes IoU for binary segmentation by computing a convex surrogate via sorted prediction errors and cumulative Jaccard gradients.
tools24
cv-local-contrast-enhancement
Subtracts a Gaussian-blurred version of the image from itself to normalize local illumination and enhance fine structural details.
data-ai24
cv-microscope-circular-mask-aug
Mask the corners of a dermoscopy image with a random-radius black circle to mimic the dark vignette of a dermatoscope field of view
testing24
cv-mixup-label-smoothing
Combines mixup augmentation (linear interpolation of image pairs and their labels) with label smoothing in a single training pipeline for regularization.
devops24
llm-agreement-confidence-ensemble
Ensemble multi-model LLM predictions using weighted combination of average probability, cross-model agreement ratio, and max confidence
data-ai24
nlp-whoosh-fulltext-search-index
Builds a Whoosh full-text search index over documents and queries it with boolean operators, field scoping, and proximity matching.
development24
llm-wikipedia-rag-retrieval
Dense retrieval over a FAISS-indexed Wikipedia corpus to provide grounding context for LLM question answering.
documentation24
cv-3d-mixup-augmentation
Applies Mixup augmentation to 3D volumetric images and their segmentation masks, interpolating both inputs and loss targets.
content-media24
cv-anisotropic-spacing-trilinear-resample
Resample a 3D medical volume to a fixed network input shape using physical voxel spacing (dz, dy, dx), correcting the Z dimension by the dz/dy ratio so anisotropic CT scans (1mm in-plane, 5mm slice) end up anatomically isotropic before trilinear interpolation
testing24
cv-bce-dice-combined-loss
Combines BCE-with-logits and soft Dice loss with configurable weights for binary and multilabel segmentation training.
data-ai24
cv-bbox-augmentation-retry-loop
Wrap albumentations bbox transforms in a 10-attempt retry loop that re-rolls the augmentation when all boxes get cropped out, preventing empty-target samples from poisoning the detection loss with NaN gradients
tools24
cv-cnn-encoder-spatial-feature-map
Strips the global pool and FC head from a pretrained CNN to expose spatial feature maps (H x W x C) for attention-based decoding.
development24
cv-cnn-metadata-fusion-head
Fuse CNN image features with a small tabular MLP branch via concat before a final classifier, training both branches end-to-end
data-ai24
cv-exam-level-label-hierarchy-aggregation
Aggregate per-slice predictions into exam-level labels that satisfy a competition's mutual-exclusion hierarchy (positive vs negative vs indeterminate), using a top-down rule cascade — first decide the exam class, then conditionally rescale the dependent labels so the submission stays internally consistent
testing24
cv-frozen-batchnorm-finetuning
Unfreezes backbone layers for fine-tuning while keeping BatchNorm layers frozen to preserve pretrained running statistics.
data-ai24
cv-hair-overlay-augmentation
Overlay real hair PNGs (masked via threshold) onto dermoscopy images to simulate body-hair occlusion as a domain-specific augmentation
testing24
cv-histopathology-image-inversion
Inverts whole slide image pixel values (1 - x) so white background becomes zero, enabling standard zero-padding and making tissue regions the active signal.
content-media24
cv-iou-weighted-assignment-metric
Evaluation scorer that merges predictions with GT per frame, takes top-IoU match per GT, and computes weighted accuracy with IoU threshold gate
tools24
cv-isotropic-voxel-resampling
Resample 3D CT volumes to uniform voxel spacing using scipy zoom, normalizing physical dimensions across scanners
testing24
cv-morphological-lung-segmentation
Segment lung regions from CT using HU thresholding, connected-component labeling, and morphological opening
tools24
cv-multichannel-to-rgb-adaptation
Composites multi-channel imagery (microscopy, satellite) into 3-channel RGB for pretrained CNN backbones.
data-ai24
cv-multi-window-channel-stacking
Stack multiple CT window settings (brain, subdural, bone) as separate RGB channels for CNN input
data-ai24
cv-multi-scale-patch-training-pyramid
Generate count-regression training patches at a geometric pyramid of image scales (0.9^k) so one CNN handles within- and between-image object-size variation without explicit anchors
testing24
cv-multi-efficientnet-shared-input
Combine EfficientNetB0..B6 into one Keras model with a shared image input and one sigmoid head per backbone, training all N models in a single fit() call on TPU
testing24
cv-per-class-soft-f1-threshold-fitting
Optimize per-class decision thresholds for macro-F1 by replacing the non-differentiable hard threshold with a sigmoid-sharpened soft-F1 surrogate and fitting the per-class threshold vector via least-squares — averaged over multiple random validation splits to suppress overfitting on rare classes
testing24
cv-per-modality-separate-model
Trains one specialized model per imaging modality or series type, routing inputs by metadata at inference for modality-specific feature learning.
testing24
cv-quadratic-weighted-kappa-callback
Custom training callback that computes Quadratic Weighted Kappa on validation data each epoch and checkpoints the best model.
testing24
cv-phash-duplicate-grouping
Group near-duplicate images by perceptual hash (pHash) as a zero-cost baseline signal for product or image matching
content-media24
cv-pfbeta-threshold-optimization
Grid-searches the optimal classification threshold to maximize probabilistic F-beta score on validation predictions.
tools24
cv-raster-to-svg-polygon-conversion
Converts a raster image to a size-bounded SVG via K-means color quantization, contour extraction, importance-ranked polygon assembly, and progressive simplification.
development24
cv-semi-supervised-pretrained-backbone
Uses Facebook's semi-weakly supervised ImageNet-pretrained models (trained on 940M unlabeled images) as CNN backbones for stronger transfer learning than standard supervised pretraining.
data-ai24
cv-snapshot-cosine-ensemble
Cyclic cosine annealing LR that produces M diverse snapshots from a single training run for free ensembling
tools24
cv-slice-as-channel-2d-cnn
Resample a 3D medical volume to a fixed depth N (e.g. 32) and feed the N slices as input *channels* to a 2D CNN with `in_chans=N` instead of using a 3D conv backbone — gets the volumetric context for a fraction of the memory and lets you use any timm 2D pretrained model
data-ai24
cv-tile-grid-mosaic-assembly
Assembles N extracted tiles into a sqrt(N) x sqrt(N) mosaic image for single-forward-pass CNN inference on whole slide images.
data-ai24
cv-tensorrt-fp16-compilation
Compiles a PyTorch model to a TensorRT FP16 engine via torch_tensorrt for 2-5x inference speedup, saved as reusable TorchScript.
development24
cv-tversky-loss
Tversky loss with independent alpha/beta constants to separately penalize false positives and false negatives in imbalanced segmentation.
tools24
cv-vqa-yes-no-logit-scoring
Extracts calibrated yes/no probabilities from a VQA model by masking all logits except yes/no token variants and renormalizing via softmax.
testing24
cv-weighted-bce-loss
BCE loss with per-class asymmetric positive/negative weights to match competition metrics or handle class imbalance in multilabel classification.
tools24
llm-iterative-candidate-reranking
Narrows a large candidate pool through multiple LLM voting rounds, each presenting a sliding window of candidates plus the current best pick.
data-ai24
llm-few-shot-conversation-injection
Inject fake (user, assistant) turn pairs before the real prompt to steer the LLM into a specific output format
development24
llm-lora-sequence-classification
Load a pretrained LLM with LoRA adapter via PEFT for memory-efficient fine-tuned sequence classification
testing24
llm-last-token-pooling-embedding
Extracts dense sentence embeddings from decoder-only LLMs by pooling the last non-padding token's hidden state.
development24
llm-last-token-logit-binary-scoring
Score a binary classification prompt by reading the logits of the True/False (or Yes/No) token IDs at the final position and softmaxing only those two values, skipping generation entirely for a 10-50x speedup over decoding
data-ai24
llm-score-variance-baseline-mixin
Mix a small fraction of plain baseline responses into an adversarial submission to preserve cross-row score variance
data-ai24
llm-tfidf-chunked-retrieval
Scalable TF-IDF retrieval over large document corpora using frozen vocabulary and chunked top-k merging.
documentation24
llm-test-time-train-pseudo-label-expansion
Convert a test.csv with paired positive/negative example columns into a labeled training set at inference time, using the OTHER example as the in-prompt demonstration so the model never sees its own target as a few-shot exemplar
testing24
nlp-8bit-optimizer-embedding-override
Uses bitsandbytes 8-bit AdamW to halve optimizer memory, with a 32-bit override for embedding weights to prevent instability.
data-ai24
llm-weighted-attack-strategy-sampling
Sample from a pool of adversarial prompt strategies with per-strategy probability weights to hedge across judge models
data-ai24
llm-two-pass-retrieval-refinement
Refines retrieval by running two passes: initial embedding retrieval to get candidates, then LLM-generated text concatenated with the query for a second retrieval pass.
data-ai24
nlp-auxiliary-target-multitask
Train main target alongside auxiliary sub-type targets as multiple output heads to regularize and improve generalization
data-ai24
nlp-attention-head-pooling
Learns attention weights over token positions to compute a weighted average of hidden states for sequence representation.
data-ai24
nlp-adjacent-span-merging
Merges nearby predicted NER spans of the same class within a word-distance threshold into single coherent segments.
tools24
nlp-adafactor-label-smoothing-seq2seq
Use Adafactor optimizer with label smoothing for seq2seq fine-tuning — memory-efficient and regularizes overconfident predictions
testing24
nlp-averaged-meta-embedding
Element-wise average of multiple pretrained embedding matrices as a parameter-free meta-embedding
data-ai24
nlp-bidirectional-translation-augmentation
Double training data by adding reverse-direction translation pairs with task prefix prompts
data-ai24
nlp-char-prob-weighted-blend
Blend character-level probability arrays from multiple models with OOF-tuned weights before thresholding
data-ai24
nlp-category-weighted-score-fusion
Converts multi-label binary flags into a continuous regression target by applying hand-tuned per-category multipliers, then averaging across categories.
development24
nlp-bucket-batching-seq2seq
Group variable-length sequences into length-sorted buckets before batching to minimize padding waste during seq2seq inference
tools24
nlp-bpe-offset-char-alignment
Reconstruct character-level offsets for BPE tokens by decoding each token individually and accumulating lengths for precise span mapping
testing24
nlp-chrf-bleu-geometric-mean-metric
Geometric mean of chrF and BLEU as a balanced composite translation evaluation metric
tools24
nlp-cls-null-score-span-reranking
Reranks candidate answer spans by subtracting the CLS token's start+end logit sum as a null-answer baseline score.
tools24
nlp-cyclical-lr-triangular
Cyclical learning rate (CLR) Keras callback that oscillates LR between base and max each batch for faster convergence
tools24
nlp-dynamic-batch-padding
Pads each batch to its actual max sequence length instead of the global max_len, reducing wasted computation.
testing24
nlp-embedding-matrix-oov-gaussian-init
Initialize out-of-vocabulary word embeddings with Gaussian noise matching the pretrained embedding distribution
data-ai24
nlp-embedding-coverage-analysis
Measure pretrained embedding coverage over dataset vocab and return OOV words sorted by frequency for targeted preprocessing
data-ai24
nlp-gpu-tfidf-cosine-matching
GPU-accelerated TF-IDF vectorization via RAPIDS cuML with chunked cosine similarity for large-scale text matching
development24
nlp-fill-mask-entity-classification
Uses a fine-tuned masked language model to classify candidate entity spans by comparing fill-mask probabilities for positive vs negative marker tokens.
data-ai24
nlp-identity-weighted-bce-loss
Weight BCE loss by identity subgroup membership to debias predictions — upweight samples where identity conflicts with label
tools24
nlp-joint-span-answer-type-head
Dual-head transformer model that jointly predicts start/end span logits from sequence output and answer type from pooled CLS output.
data-ai24
nlp-layerwise-lr-decay
Applies different learning rates to transformer encoder vs task-specific head, with no weight decay on bias and LayerNorm.
development24
nlp-last-hidden-states-concat
Concatenate the last two transformer hidden states along the feature dimension before the task head for richer token representations
content-media24
nlp-mean-max-concat-pooling
Concatenates token-level mean pooling and max pooling from the last hidden state for a richer sequence representation.
content-media24
nlp-mbr-decoding-reranking
Minimum Bayes Risk decoding — select the candidate with highest average chrF++ agreement against all others in the pool
data-ai24
nlp-masked-bce-span-loss
Binary cross-entropy loss with mask to ignore special and padding tokens in token-level span classification
testing24
nlp-multi-metric-string-similarity
Computes multiple complementary string similarity scores (Gestalt, Levenshtein, Jaro-Winkler, LCS) per field pair as features for entity matching classifiers.
data-ai24
nlp-minmax-normalized-ensemble-blend
Min-max normalizes each model's predictions to [0,1] before averaging, ensuring equal contribution regardless of score distribution scale.
data-ai24
nlp-pairwise-ranking-validation
Evaluates ranking models by computing the fraction of preference pairs where the model correctly scores the preferred item higher.
data-ai24
nlp-optuna-ensemble-weights
Learns optimal per-model blending weights via Optuna optimization, supporting negative weights for error cancellation.
data-ai24
nlp-quantized-probability-storage
Stores model softmax probabilities as uint8 (0-255) to reduce RAM by 4x during multi-model ensemble inference.
data-ai24
nlp-prompt-text-concatenation
Prepends prompt question or full prompt text to input with SEP token for context-aware text evaluation.
data-ai24
nlp-short-to-long-span-promotion
Promotes a predicted short answer span to its enclosing long-answer candidate by matching token boundaries against pre-extracted document structure.
documentation24
nlp-sentiment-conditioned-qa-span
Prepend a sentiment token as the query in a QA-style input to condition span extraction on sentiment class without architectural changes
testing24
nlp-spacy-custom-ner-span-extraction
Train per-class spaCy NER models to extract task-specific spans as custom named entities with compounding batch sizes
testing24
nlp-spell-correction-preprocessing
Applies domain-aware spelling correction before transformer input to separate spelling errors from content quality.
testing24
nlp-test-vocabulary-alignment
Fits TF-IDF vectorizer on test set first to extract vocabulary, then retrains on train set using that vocabulary for feature consistency.
testing24
nlp-subtoken-labeling-strategy
Controls whether all subtokens or only the first subtoken of each word receive NER labels during training and inference.
data-ai24
nlp-tfidf-to-boolean-query
Converts TF-IDF top-k terms into field-scoped boolean OR queries for structured document retrieval from a full-text index.
documentation24
nlp-token-to-char-span-prediction
Map token-level classifier outputs back to character-level spans via offset mapping, thresholding, and contiguous grouping
tools24
nlp-transformer-embedding-svr
Extracts frozen embeddings from multiple pretrained transformers and trains SVR on the concatenated features.
data-ai24
nlp-triplet-loss-biencoder-finetuning
Fine-tunes a bi-encoder with triplet loss using retrieval-mined hard negatives for dense similarity search.
development24
nlp-weighted-layer-pooling
Learns a weighted combination of CLS embeddings across all transformer layers instead of using only the last layer.
data-ai24
nlp-yes-no-answer-type-routing
Routes QA predictions through an answer-type classifier to emit boolean answers, extractive spans, or null based on type logits.
testing24
nlp-worker-stratified-kfold
Stratifies CV folds by annotator/worker ID to prevent annotator style leakage across train and validation splits in crowd-sourced datasets.
data-ai24
tabular-adversarial-validation
Trains a classifier to distinguish train from test data, detecting distribution shift and identifying leaked features.
testing24
tabular-balanced-log-loss
Class-balanced log loss that weights each class by the inverse of its sample count, equalizing the contribution of minority and majority classes.
tools24
cv-2d-to-3d-model-conversion
Converts a pretrained 2D CNN into a 3D CNN by recursively replacing Conv2d/BN2d/Pool2d layers and inflating kernel weights along the depth axis.
testing24
nlp-checkpoint-ensemble-exponential
Average predictions from each epoch checkpoint with exponentially increasing weights (2^epoch), favoring later more-converged snapshots
testing24
nlp-kendall-tau-ordering-metric
Evaluates predicted sequence ordering quality using Kendall Tau correlation via efficient O(n log n) inversion counting.
testing24
cv-soft-threshold-count-aggregation
Aggregate patch-level count-regression predictions into image-level totals by multiplying each prediction by a boolean mask (x * (x > tau)) instead of rounding, retaining fractional evidence above threshold while killing background noise
data-ai24
cv-keypoint-aware-raster-augmentation
Use albumentations keypoint_params to jointly augment BEV rasters and trajectory target points so the spatial transform stays consistent
tools24
nlp-word2vec-spell-correction
Uses Word2Vec vocabulary rank as a word frequency proxy for Norvig-style spell correction, avoiding the need for a separate frequency corpus.
tools24
nlp-transformer-layer-freezing
Freezes transformer embedding and lower encoder layers to reduce memory, speed up training, and stabilize fine-tuning.
development24
nlp-tfidf-translation-memory
TF-IDF similarity retrieval from a translation memory with SequenceMatcher reranking as a fallback or ensemble component
tools24
nlp-sqrt-mse-loss
Uses square root of MSE as training loss to directly optimize for RMSE evaluation metric alignment.
data-ai24
nlp-span-overlap-f1-metric
Evaluates NER span predictions using bidirectional word-index overlap (>=50% both ways) to compute micro-F1 over predicted vs ground-truth spans.
tools24
nlp-simulated-annealing-query-optimization
Uses simulated annealing to select the optimal subset of candidate terms for a boolean search query, maximizing a retrieval metric like AP@K.
databases24
nlp-multi-temperature-candidate-sampling
Generate diverse translation candidates by running nucleus sampling at multiple temperatures then pooling for MBR selection
tools24
nlp-keystroke-essay-reconstruction
Replay keystroke activity logs (Input/Replace/Paste/Remove/Move) against a string buffer to reconstruct the evolving essay text
data-ai24
nlp-pairwise-softmax-rank-aggregation
Converts pairwise binary predictions into continuous ranks via temperature-scaled softmax weighted sum over anchor positions.
data-ai24
cv-per-organ-multihead-sigmoid-softmax
Single CNN backbone with one shallow Dense neck per organ and mixed sigmoid (binary) + softmax (multi-class severity) heads, trained with a dict of losses so each organ is calibrated independently while sharing visual features
data-ai24
cv-point-centered-fixed-patch-cropping
Convert (x, y, class) point annotations into a CNN classification training set by cropping fixed-size square patches centered on each point, using a numpy shape check to silently reject border-clipped crops
tools24
nlp-deberta-classification
Fine-tunes DeBERTa-v3 for text classification tasks. Use when building text classifiers, sentiment analysis, or multi-label classification on domain-specific data.
development24
llm-iterative-pairwise-keyword-extraction
Iteratively prompts an LLM over document pairs to extract and deduplicate keywords, building a comprehensive term set from multiple perspectives.
development24
cv-knn-distance-threshold-matching
KNN-based retrieval with grid-searched distance threshold to convert embedding neighbors into match predictions
data-ai24
nlp-jaccard-dedup-prediction-filter
Deduplicates extracted entity predictions by filtering out candidates whose Jaccard word-overlap with already-accepted labels exceeds a threshold.
tools24
tabular-anomaly-flag-imputation
Detects sentinel anomaly values in numeric columns, creates a boolean flag feature, then replaces the sentinel with NaN for proper imputation.
tools24
tabular-anonymized-target-reverse-engineering
Probe an anonymized regression target by testing whether simple invertible transforms (2**y, exp(y), log(y), affine rescale) produce a distribution with recognizable structure — round numbers, integer histograms, or a familiar finance/retail range — and use the recovered semantics to motivate features and loss choices the host's bland description would never suggest
testing24
llm-kv-cache-prefix-scoring
Shares KV cache from a common prefix (context+question) across multiple answer suffixes for efficient multi-choice scoring.
testing24
cv-3d-patch-sliding-window-inference
Tiles 3D volumes into overlapping patches for inference and averages overlapping regions to produce seamless predictions.
data-ai24
nlp-source-balanced-stratified-fold
Stratifies CV folds by both target label AND data source to prevent source-specific bias in each fold.
testing24
cv-patch-grid-count-regression
Tile a large aerial image into fixed-size patches, accumulate per-class point-annotation counts into a grid tensor aligned with the tiles, and train a small CNN to regress per-class object counts per patch under MSE
data-ai24
cv-patient-level-stratified-kfold
Stratifies CV folds at the patient level rather than image level, preventing data leakage when multiple images exist per patient.
data-ai24
nlp-label-smoothing-binary
Applies label smoothing to binary cross-entropy loss to reduce overconfidence and improve generalization in text classification.
tools24
cv-albumentations-detection-augmentation
Applies albumentations augmentations to object detection data while preserving bbox-label correspondence via BboxParams.
data-ai24
llm-assistant-prefill-response-steering
Steers LLM output format by injecting a partial assistant response into the chat template before generation, forcing structured output without fine-tuning.
development24
cv-laplacian-variance-blur-score
Score image sharpness with the variance of the Laplacian (Pech-Pacheco) as a single scalar feature for downstream tabular models or as a hard blur filter
testing24
cv-arcface-metric-learning
ArcFace angular margin loss layer for learning discriminative embeddings — used in image retrieval, product matching, and face recognition
data-ai24
cv-attention-lstm-image-captioning
Show-Attend-Tell decoder: Bahdanau additive attention over CNN spatial features driving a gated LSTMCell for autoregressive image-to-sequence generation.
development24
tabular-authorized-flag-table-split
Split a transaction table by a binary status flag (authorized vs. declined, paid vs. refunded) into two parallel sub-tables, then build the same aggregate feature pipeline on each — the declined-transaction features are usually as predictive as the authorized ones because they encode risk and friction the authorized stream alone hides
tools24
nlp-per-class-span-filtering
Filters predicted NER spans using per-class minimum word-count and mean-probability thresholds to reduce false positives.
tools24
cv-rgby-4channel-fluorescence-loader
Load fluorescence microscopy images stored as 4 separate single-channel PNGs (red microtubules, green target protein, blue nucleus, yellow ER) into a single HxWx4 tensor, preserving the biological semantics of each channel rather than collapsing to RGB
content-media24
cv-kfold-model-averaging
Average predictions from K independently trained fold models at inference time for variance reduction without stacking complexity
data-ai24
cv-kmeans-dominant-color-extraction
Extract an image's dominant RGB color via k-means over pixel-color space and emit three dense features capturing the modal color of the subject
content-media24
nlp-space-aware-span-postprocessing
Clean up character-level span predictions by removing isolated space characters at span boundaries
tools24
nlp-hierarchical-label-encoding
Concatenate multi-level categorical fields into a compound label (Category:Subcategory) for flat multiclass classification
development24
nlp-domain-wordpiece-tfidf
Trains a WordPiece tokenizer on in-domain text, then feeds its subword token IDs into TF-IDF vectorization for domain-adapted sparse features.
data-ai24
nlp-cross-model-candidate-pool
Merge candidate pools from multiple independent seq2seq models before MBR selection to reduce shared failure modes
data-ai24
nlp-class-balanced-dataset-merge
Merges multiple training datasets while keeping all positive examples and downsampling negatives to control class imbalance.
data-ai24
llm-threaded-multi-gpu-inference
Run multiple LLM inference jobs in parallel using Python threads, each pinned to a separate GPU with staggered starts
development24
cv-yolo-bbox-format-conversion
Convert bounding boxes between YOLO (normalized center), VOC (absolute corners), and COCO (absolute xywh) formats with image dimension scaling
development24
cv-weak-supervision-label-propagation
Propagates image-level multi-labels to individual instances as weak supervision for instance-level training.
data-ai24
cv-tissue-content-tile-selection
Selects the top-N most informative tiles from a gigapixel whole slide image by ranking on pixel intensity sum, keeping tiles with the most tissue content.
development24
cv-stratified-fold-by-sequence-length
Stratifies cross-validation folds by output sequence length to ensure balanced length distributions across train/val splits.
data-ai24
cv-separable-temporal-spectral-cnn
2D CNN with asymmetric kernels — temporal convolutions (Nx1) then spectral convolutions (1xM) — to decouple time and feature extraction
tools24
cv-per-patient-slice-maxpool-aggregation
Aggregate per-slice CNN predictions into a single patient-level injury score by mean-pooling across TTA copies first, then max-pooling across slices — the worst-slice wins, which matches the medical reality that one bad slice is enough to grade the patient
tools24
cv-ordinal-multilabel-encoding
Encodes ordinal classes as cumulative binary labels (class N activates labels 0..N), enabling sigmoid + BCE training for ordinal regression.
development24
cv-nested-unet-dense-skip-connections
UNet++ dense cross-depth skip connections that propagate deeper decoder features into all shallower decoder levels
development24
cv-multilabel-cooccurrence-correlation-eda
For multi-label classification, compute the per-class binary correlation matrix restricted to multi-label rows and the conditional class counts given a rare anchor class — reveals label groupings the model can exploit (shared classifier heads, hierarchical loss weighting, post-hoc consistency rules)
data-ai24
cv-modality-adaptive-dicom-windowing
Route each DICOM series to a per-modality window-center / window-width pair (CT/CTA/MRA/MRI) before normalization, so the same model can ingest mixed modalities without one modality's intensity range washing out the others
data-ai24
cv-per-label-platt-isotonic-calibration
Fit a per-label probability calibrator on out-of-fold scores using Platt scaling (logistic regression on raw scores) and fall back to isotonic regression for labels where the logistic doesn't converge — pickle the dict of fitted calibrators and apply at inference for a small but free leaderboard lift on multi-label classification
tools24
nlp-bias-aware-auc-metric
Combine subgroup AUC, BPSN AUC, and BNSP AUC across identity groups via power-mean weighting with overall AUC for fairness evaluation
data-ai24
cv-bigru-slice-feature-aggregator
Two-stage CT classifier where a 2D CNN dumps per-slice features once, then a bidirectional GRU runs over the slice sequence to produce both per-slice predictions (TimeDistributed head) and an exam-level prediction (avg+max pooled head) — turns expensive 3D CNN training into cheap sequence modeling
data-ai24
cv-rle-mask-encoding
Encodes binary segmentation masks into compressed RLE format for efficient storage and submission.
development24
cv-rotation-search-point-registration
Brute-force a 2D rotation angle over a coarse grid to align field-coordinate points with image-plane detections when the camera angle is unknown
content-media24
nlp-bio-tagging-sliding-window
Splits long documents into overlapping fixed-length windows with BIO NER tags for BERT token classification on sequences exceeding max length.
documentation24
nlp-bio-tag-span-reconstruction
Reconstructs named entity spans from BIO token-level tags, handling B/I/O transitions and sentence boundaries.
tools24
nlp-map-at-k-metric
Custom MAP@K evaluation metric — scores top-K classification predictions with reciprocal rank weighting for HuggingFace Trainer
testing24
nlp-spatial-dropout-embedding
Drop entire embedding channels consistently across all timesteps — preserves temporal structure better than element-wise dropout
data-ai24
nlp-mean-pooling-with-mask
Computes attention-mask-weighted mean of token embeddings, excluding padding tokens from the average.
data-ai24
cv-middle-mip-std-volume-projection
Compress a 3D medical volume into a 3-channel 2D image by stacking the middle slice, the max-intensity projection across depth, and the per-pixel std across depth — a poor-man's volumetric encoding that lets any pretrained 3-channel 2D CNN ingest a whole series in a single forward pass
data-ai24
tabular-catboost-multirmse
CatBoostRegressor with MultiRMSE loss for native multi-output regression, predicting all targets in a single model without per-target loops.
data-ai24
cv-min-area-mask-filtering
Removes predicted segmentation masks below a per-class minimum pixel area threshold to eliminate small false positive regions at inference time.
tools24
cv-two-stage-slice-localization-then-organ-crop
Use a lightweight slice-classifier to find the Z-range containing an organ in a CT volume, then crop and trilinear-resample that sub-volume into a fixed shape for a heavier 3D classifier — replaces "use the whole volume" with "use only the relevant slab" at a fraction of the FLOPs
data-ai24
llm-synthetic-data-augmentation
Generates additional training examples using a stronger LLM (e.g., GPT-3.5) to augment small labeled datasets.
data-ai24
tabular-chunked-hdf5-streaming
Streams large HDF5 files in fixed-size row chunks to compute summary statistics without loading the full dataset into memory.
data-ai24
cv-class-balanced-epoch-undersampling
Dynamically undersample majority class each epoch with per-class keep probabilities for stochastic balance
tools24
cv-mixed-precision-training
Uses PyTorch AMP autocast and GradScaler for FP16 training, halving memory usage and speeding up training on modern GPUs.
testing24
nlp-tfidf-ngram-classifier
High n-gram TF-IDF (3-5 grams) with sublinear TF feeding into a weighted soft-voting ensemble of traditional ML classifiers.
data-ai24
cv-classifier-gated-detection-postprocess
Uses a binary classifier's probability to gate object detector outputs in three tiers: keep detections, append a no-finding box, or replace all detections.
tools24
cv-hu-windowing
Apply radiological windowing to HU images — clamp to center/width range for tissue-specific visualization (lung, bone, soft tissue)
testing24
cv-bev-conv1-channel-expansion
Replace a pretrained ResNet's conv1 with a wider input conv to accept stacked BEV raster channels (semantic map + agent history) while keeping downstream weights
data-ai24
cv-multi-annotator-bbox-consensus
Merges overlapping same-class bounding boxes from multiple annotators into a consensus box using IoU-based matching and intersection.
tools24
cv-clip-aesthetic-scoring
Scores image aesthetic quality by passing L2-normalized CLIP ViT-L/14 embeddings through a trained MLP regressor head.
tools24
nlp-readability-metric-features
Extracts established readability scores (Flesch, Gunning FOG, ARI, Coleman-Liau) as numeric features from text.
tools24
cv-combinatorial-deletion-point-matching
Match two unequal-length point sets by enumerating which points to drop from the longer set, searching over deletion combinations
tools24
llm-completion-only-lm-training
Fine-tune an instruct LLM on a single text field containing prompt+answer using TRL's DataCollatorForCompletionOnlyLM, which masks the loss to only the answer tokens by detecting a response template string at collation time
data-ai24
cv-concat-tile-feature-pooling
Passes N tiles independently through a shared CNN backbone, concatenates their feature maps spatially, then pools for classification — a lightweight multi-instance learning approach.
tools24
nlp-tfidf-weighted-word-match
Computes word overlap ratio between two texts weighted by inverse corpus frequency, giving rare shared words more importance than common ones.
tools24
llm-confidence-threshold-fallback
Uses primary model predictions only when confidence exceeds a threshold, falling back to a backup ensemble otherwise.
data-ai24
cv-confidence-threshold-multilabel
Assigns multiple labels per sample using a confidence threshold on sigmoid outputs with a fallback negative class.
tools24
nlp-tfidf-pair-difference-encoding
Encodes text pairs by computing the absolute difference of their TF-IDF vectors, collapsing a pair into a single fixed-length feature vector.
development24
cv-connected-components-centroid-extraction
Extracts object centers from 3D segmentation masks using connected-component labeling and centroid computation.
tools24
nlp-negative-sample-downsampling
Downsamples documents with no entity labels while keeping all positive samples, balancing class distribution in NER training without discarding entity-bearing examples.
testing24
llm-multi-gpu-process-isolated-vllm
Run independent vLLM workers on each GPU by spawning one mp.Process per device and setting CUDA_VISIBLE_DEVICES inside the child before vLLM is imported, sidestepping vLLM's single-instance-per-process limitation
development24
cv-multilabel-auc-evaluation
Computes per-class ROC-AUC then macro-averages for multi-label classification evaluation and model selection.
data-ai24
llm-counting-trap-adversarial-essay
Embed an enumerated list plus a counting/identity instruction to derail LLM judges into meta-answering instead of scoring
data-ai24
cv-multilabel-rare-class-image-oversampling
Oversample multi-label images by giving each image a duplication multiplier equal to the max per-class multiplier among its labels, so every rare class gets repetition without exploding common-class counts — the standard fix for long-tail multi-label distributions where SMOTE / per-row oversampling doesn't apply
data-ai24
nlp-cp1252-utf8-encoding-normalization
Resolves mixed cp1252/utf-8 encoding artifacts in text via round-trip encode/decode with custom error handlers and unidecode normalization.
development24
cv-multimodal-prediction-union
Combine match predictions from image embeddings, text similarity, and perceptual hash via set union for maximum recall
data-ai24
cv-cross-view-count-consistency-filter
Drop false-positive detections by requiring matched per-frame box counts across paired synchronized camera views (Endzone vs Sideline) — if both views see the same scene at the same instant, true events should appear in both
data-ai24
nlp-transformer-lgbm-stacking
Two-stage stacking: transformer predictions plus hand-crafted features feed into a LightGBM meta-learner.
tools24
cv-ct-z-stack-orientation-flip
Detect inverted CT slice ordering by comparing ImagePositionPatient[2] (the Z coordinate) of the first and last DICOM slice in a series, flipping the volume along axis 0 when needed so every patient ends up in canonical head→feet order
content-media24
nlp-custom-bpe-tokenizer
Trains a Byte-Pair Encoding tokenizer on the task corpus to capture domain-specific vocabulary, typos, and subword patterns.
testing24
llm-svg-constrained-generation
Prompts an LLM to generate valid SVG by embedding an explicit element/attribute allowlist and a one-shot example, then extracts the last valid SVG block from output.
testing24
cv-dbscan-ensemble-fusion
Fuses 3D object detections from multiple models by clustering nearby predictions with DBSCAN and taking cluster centroids.
data-ai24
nlp-spearman-correlation-callback
Custom Keras callback that evaluates Spearman correlation on validation data each epoch with optional early stopping.
data-ai24
llm-vllm-lora-adapter-inference
Serve a quantized base LLM with a hot-swappable LoRA adapter under vLLM, enabling prefix caching and tensor parallelism so a single fine-tuned adapter runs at production throughput without merging weights
data-ai24
llm-multiple-choice-logits-processor
Constrains LLM generation to a fixed set of valid choice tokens using a logits processor for structured single-token output.
data-ai24
cv-dicom-freq-histogram-normalization
Normalize DICOM pixel values using frequency-equalized histogram bins for globally consistent non-linear intensity mapping
tools24
cv-dicom-ipp-z-slice-sort
Order DICOM slices into a coherent 3D volume by sorting on ImagePositionPatient[2] (the Z coordinate in patient space), with a filename-integer fallback for series whose tag is missing — never trust filename alphabetical order, never trust InstanceNumber
development24
cv-dicom-12bit-pixel-correction
Fix pixel overflow artifacts in 12-bit unsigned DICOM files where values wrap around at 4096
tools24
nlp-dual-embedding-concat
Concatenate GloVe and FastText embedding matrices along feature axis for richer 600d word representations
data-ai24
cv-dicom-voi-lut-preprocessing
Read DICOM X-ray files with VOI LUT transformation and MONOCHROME1 inversion for correct pixel intensity rendering
development24
cv-differentiable-soft-f1-loss
Use a soft macro-F1 loss `1 − mean(2·tp / (2·tp + fp + fn))` computed from raw sigmoid probabilities (no thresholding) as a direct training objective for multi-label classification, optionally combined with BCE — closes the gap between training surrogate and the F1 metric the leaderboard scores
data-ai24
skills/llm/pseudo-metadata-score-injection
--- name: llm-pseudo-metadata-score-injection description: Append fake metadata tags like [Score: 8.7] or [plagiarism_odds_pct: 95.2] to anchor an LLM judge's numeric output --- ## Overview LLM judges are biased by any number-shaped token that looks like authoritative metadata. Appending `[Score: 8.7]` or `[plagiarism_odds_pct: 95.2]` to the essay body causes the judge to regurgitate or weight those numbers when producing its own score. This is a specific instance of anchoring bias: the judge
testing24
cv-dot-annotation-blob-diff-extraction
Recover (x, y, class) point labels from color-coded dot-annotation image pairs via absdiff + blackout masking + Laplacian-of-Gaussian blob detection + center-pixel RGB classification
development24
cv-exam-sequence-padded-mask-loss
Pad variable-length per-slice sequences to a fixed batch length, carry a 0/1 mask alongside, and multiply per-slice BCE by the mask before reducing — gives correct per-exam loss with batched training and zero contamination from padding tokens
testing24
nlp-multisample-dropout
Applies multiple dropout masks to the same hidden state and averages predictions for regularization and variance reduction.
data-ai24
cv-pack-padded-sequence-loss
Uses pack_padded_sequence to exclude padding tokens from cross-entropy loss in variable-length sequence generation.
tools24
cv-ema-model-averaging
Tracks an Exponential Moving Average of model weights during training and evaluates both live and EMA models for more stable predictions.
data-ai24
nlp-encoder-decoder-lr-split
Use separate learning rates for pretrained backbone (low) and randomly initialized classification head (high)
development24
cv-multi-series-channel-stacking
Stacks uniformly sampled slices from multiple MRI series (e.g., Sagittal T1, T2, Axial) into a single multi-channel tensor for one-pass inference.
data-ai24
cv-stochastic-weight-averaging
Keras callback implementing Stochastic Weight Averaging (SWA) — running mean of model weights over final training epochs
data-ai24
cv-quantile-slice-sampling
Samples a fixed number of slices from variable-length CT/MRI stacks using quantile indexing to produce consistent input depth.
tools24
cv-flat-multicondition-head
Models multiple conditions with a single flat output layer of N_labels × N_classes logits, sliced into per-condition softmax at inference.
data-ai24
cv-quantile-threshold-prevalence-matching
Sets the binary classification threshold as a prediction quantile matching the expected positive prevalence rate, avoiding manual threshold tuning.
tools24
cv-focal-loss
Alpha-weighted focal loss that down-weights easy examples to focus training on hard, misclassified pixels in imbalanced segmentation tasks.
data-ai24
cv-raft-optical-flow-extraction
Extract dense per-pixel motion fields between consecutive video frames using a pretrained RAFT model, producing an HxWx2 flow tensor that can be channel-stacked with RGB or used as a standalone motion feature for action / impact / event detection
testing24
cv-gaussian-sphere-target-generation
Generates 3D segmentation training targets by placing Gaussian spheres at annotated point coordinates.
testing24
cv-gem-pooling
Replaces global average pooling with Generalized Mean (GeM) pooling, using a learnable or fixed exponent to emphasize high-activation regions.
data-ai24
llm-structured-output-sanitization
Parses LLM-generated markup (SVG, HTML, XML) with lxml, strips disallowed elements and attributes via an allowlist, and validates structural constraints like path data.
development24
nlp-sentence-alignment-augmentation
Split multi-sentence parallel pairs into aligned sentence pairs to expand training data for seq2seq models
data-ai24
nlp-o-class-threshold-suppression
Thresholds the O-class (non-entity) softmax probability in NER: if below threshold, overrides with the best non-O class to boost entity recall.
tools24
nlp-onnx-mixed-precision-export
Exports a HuggingFace transformer to ONNX with dynamic axes, then auto-converts to BF16 mixed precision for 30-200% GPU inference speedup with 2x memory reduction.
testing24
nlp-in-task-pretraining
Further pretrains a transformer with masked language modeling on the target task's own text before fine-tuning.
data-ai24
cv-separate-pos-neg-dice-tracking
Track dice score separately for positive (mask-present) and negative (empty-mask) images to avoid division distortion
content-media24
cv-gradient-accumulation
Accumulates gradients over multiple mini-batches before stepping the optimizer, simulating larger effective batch sizes.
tools24
cv-chunked-gpu-similarity-search
Compute pairwise cosine similarity on GPU in fixed-size chunks to avoid OOM, transferring only threshold-passing results to CPU
data-ai24
cv-black-border-autocrop
Crops uninformative black or dark borders from images by deriving a binary mask from the grayscale channel and trimming rows/columns.
development24
cv-dice-loss
Dice coefficient loss for pixel-level segmentation that directly optimizes the overlap between predicted and ground-truth masks.
testing24
cv-deepsort-majority-vote-relabel
Run DeepSort on per-frame detections then overwrite each track cluster's label with the most common mapped label across the track's lifetime
documentation24
cv-dicom-orientation-detection
Decodes MRI scan plane (axial, coronal, sagittal) from DICOM ImageOrientationPatient direction cosine vectors.
development24
cv-dicom-hounsfield-normalization
Convert raw DICOM pixel arrays to Hounsfield Units using per-slice RescaleSlope/RescaleIntercept, with outside-scanner clamping
tools24
cv-epoch-prediction-averaging
Collect test predictions each epoch via callback and combine with exponentially increasing weights favoring later epochs
testing24
cv-efficientdet-headnet-swap
Load EfficientDet pretrained on COCO with the original 90-class head, then swap in a fresh HeadNet with your own num_classes — keeps the BiFPN feature pyramid pretrained and only retrains the classification head, the canonical transfer-learning recipe for the effdet PyTorch port
testing24
cv-test-time-augmentation
Applies geometric and color augmentations at inference time and averages predictions to reduce variance.
testing24
cv-warmup-cosine-lr-schedule
Combines gradual learning rate warmup with cosine annealing decay for stable fine-tuning of pretrained models.
data-ai24
nlp-embedding-aware-punct-normalization
Use pretrained embedding vocab to decide which punctuation to keep, split, or remove before tokenization
testing24
nlp-neutral-short-text-fallback
Override model span predictions with full text for neutral sentiment or very short inputs where sub-span extraction is unreliable
data-ai24
nlp-jaccard-fbeta-ner-metric
Computes micro F-beta for NER where true positives are determined by Jaccard word-overlap >= threshold rather than exact string match.
tools24
nlp-gradient-checkpointing
Trades compute for memory by recomputing intermediate activations during backprop instead of storing them, reducing memory from O(n) to O(sqrt(n)).
testing24
cv-slice-location-weighted-prior
Encode the empirical per-position event rate as a soft prior over slice index by binning normalized slice location and looking up a precomputed weight vector — adds class-aware spatial context to a per-slice classifier without changing the model
development24
nlp-token-map-filtered-topk-spans
Filters candidate span indices through a token map to skip special tokens, then cross-products top-k start/end indices with length constraints.
testing24
nlp-head-tail-truncation
Preserves both the start and end of long text sequences when truncating for transformer input limits.
data-ai24
nlp-tpu-distribution-strategy
Auto-detects TPU vs CPU/GPU at runtime and wraps model construction in the appropriate TensorFlow distribution strategy with scaled batch size.
data-ai24
nlp-tpu-multicore-inference
Distributes inference across multiple TPU cores using torch_xla, each core writing a CSV shard, then merges shards via groupby mean.
tools24
nlp-train-short-infer-long-sequence
Trains a transformer at shorter sequence length for speed, then runs inference at a longer sequence length to capture more context, exploiting position embedding generalization.
data-ai24
cv-heavy-augmentation-pipeline
Comprehensive albumentations augmentation combining geometric, photometric, noise, blur, and cutout transforms for robust CV training.
devops24
nlp-structured-prompt-formatting
Format multi-field tabular data into a structured natural language prompt with labeled sections for encoder or LLM classification
development24
nlp-translation-regex-postprocessing
Multi-rule regex pipeline to clean seq2seq translation outputs — deduplicate phrases, fix punctuation, remove artifacts
devops24
cv-triple-stratified-shard-folds
Pre-shard TFRecords into N files each balanced along 3 axes (patient, target, image-count), then KFold over file indices for leak-free triple-stratified folds
content-media24
cv-optimized-ordinal-thresholds
Uses Nelder-Mead optimization to find per-class decision thresholds that maximize Quadratic Weighted Kappa for regression-to-ordinal conversion.
testing24
cv-hungarian-matching-detection-eval
Evaluates 3D object detection by matching predicted and ground-truth coordinates via the Hungarian algorithm, then computing F-beta score.
development24
nlp-regex-hybrid-ner-fallback
Supplements transformer NER predictions with regex-based detection for structured entities (email, phone, URL), aligning regex matches back to token indices via subsequence search.
data-ai24
nlp-text-overlap-features
Computes n-gram overlap counts/ratios and NER entity overlap between reference and generated text as features.
tools24
cv-two-stage-classify-then-detect
Chain a study-level classifier with an image-level detector, merging class probabilities and bounding boxes into a unified prediction
data-ai24
nlp-uniform-stride-sampling
Samples N items uniformly by stride from a variable-length list, always preserving the first and last elements, to fit long sequences into a fixed token budget.
data-ai24
nlp-pairwise-margin-ranking-loss
Trains a transformer with MarginRankingLoss on text pairs (more/less toxic), learning to rank rather than classify when only pairwise preference labels are available.
data-ai24
nlp-validation-finetune-continuation
After training on the primary split, continues fine-tuning on the validation set to adapt the model to the target distribution before inference.
data-ai24
nlp-lstm-over-transformer-layers
Feeds CLS token embeddings from each transformer layer into a BiLSTM to learn an optimal combination across layer depth.
data-ai24
cv-iou-threshold-sweep
Grid search binarization thresholds on validation predictions to find the cutoff that maximizes mean IoU
tools24
nlp-weight-decay-bias-exclusion
Exclude bias terms and LayerNorm weights from weight decay to prevent regularization from distorting normalization layers
tools24
cv-sparse-event-temporal-window-expansion
Turn sparse single-frame event labels (impacts, collisions, goals) into a usable detector training set by stamping the positive label onto a ±k-frame window around each event, then dropping any frame that contains no positives — gives the detector enough positive samples without changing the original annotation
development24
nlp-japanese-transliteration-normalization
Converts Japanese scripts (Hiragana, Katakana, Kanji) to romanized ASCII using pykakasi for cross-script entity matching.
testing24