| # | Misreading | Reality | Check |
|---|---|---|---|
| 1 | "This cluster is bigger, so the group is more diverse" | t-SNE equalises densities; UMAP partly too | compare within-cluster distances in the original space; densMAP |
| 2 | "Clusters A and B are far apart, so they are very different" | inter-cluster distances are not preserved (t-SNE especially) | distances between cluster centroids in the original space |
| 3 | "There are 7 clusters" | perplexity / n_neighbors change the count | grid over parameters and seeds; clustering in the original space |
| 4 | "These points are outliers" / "there are no outliers" | outliers are pulled into nearby clusters (seen on the blobs fixture) or appear as fake islands | outlier detection in the original space |
| 5 | "The map shows two groups" in pure noise | small perplexity produces clumps even in random data | run on shuffled data; trustworthiness |
| 6 | "Points along this curve form a trajectory" | gradients can be cut or bent arbitrarily | colour by the variable; PCA view |
| 7 | "Model separates classes well, look at the map" | the map can separate classes the model cannot (or hide separation) | label k-NN accuracy original vs map (embedding_trust.py) |
| 8 | "The layout changed, so the data changed" | different seeds give rotated/reflected/rearranged layouts | fix seeds; compare neighbourhoods, not positions |
| 9 | "Supervised UMAP shows the labels are separable" | it used the labels | unsupervised map for evidence; supervised only for display |
| 10 | "Clusters on the t-SNE plot, let's run k-means on its coordinates" | t-SNE coordinates distort density and distance | cluster on original / PCA / 10-20-D UMAP space |
Sanity checks before presenting a map
- Same structure in at least two seeds and two parameter values.
- PCA baseline shown alongside (with explained variance).
- Trustworthiness / continuity reported (> 0.9 for strong claims).
- Any "cluster" claim confirmed by clustering in a faithful space and by profiling.
- Caption: "Positions show neighbourhoods; sizes and distances between groups are not to scale."
A null check
Shuffle each feature column independently (destroys structure), embed with the same parameters: if the map of shuffled data shows "clusters" similar to the real one, the real clusters are not evidence.
anomaly-detection-model-builder
Build and evaluate anomaly detection for tabular, sensor and transaction data with few or no labels: define the anomaly type (point, contextual, collective), choose detectors (robust z-score/MAD and IQR rules, Mahalanobis with robust covariance, k-NN distance, LOF, Isolation Forest, one-class SVM, PCA reconstruction / autoencoders, forecasting residuals for time series), combine them by rank, explain each alert by its deviating features, set thresholds from an alert budget or the score distribution, and evaluate with precision@k, PR-AUC and recall on known incidents, or detector agreement when no labels exist. Includes pure-Python detectors (robust z, Mahalanobis, kNN, Isolation Forest, ensemble) and an evaluation script.
catboost-model-builder
Build CatBoost models that use its strengths correctly: raw categorical columns as cat_features (ordered target statistics, no manual encoding), text_features, Pool construction with NaN-safe categories, ordered boosting and has_time for temporal data, symmetric (oblivious) trees, overfitting detector with eval_set and use_best_model, losses for classification, regression (Quantile, Poisson, Tweedie) and ranking (YetiRank), class weights, monotone constraints, feature importance (PredictionValuesChange, LossFunctionChange) and SHAP values, GPU training and model export. Plans cat/text/numeric columns from a CSV and generates a runnable training script, and checks parameter sets for the conflicts CatBoost rejects (iterations vs n_estimators, Bayesian bootstrap with subsample, multiple class-weight options, depth > 16, use_best_model without eval_set).
categorical-encoding-selector
Choose and implement the right encoding for every categorical column given the model family and the data: one-hot (with infrequent-level grouping and handle_unknown), ordinal with an explicit order, cross-fitted target encoding (scikit-learn TargetEncoder, CatBoost ordered statistics, category_encoders inside CV), frequency/count encoding, feature hashing, native categorical support in LightGBM, HistGradientBoosting, XGBoost and CatBoost, prefix levels for codes (zip3, NAICS-2), and embeddings. Measures cardinality, rare-level share, unseen-level rate on a time-ordered holdout and target signal per column, emits the ColumnTransformer, and quantifies target-encoding leakage (in-sample vs out-of-fold vs holdout AUC).
classification-model-selector
Choose the right classification algorithm for tabular, text or mixed data and prove the choice with a fair bake-off: logistic regression (L2, L1, elastic net), linear SVM, naive Bayes, k-NN, RBF SVM, decision trees, random forest, extra trees, HistGradientBoosting, LightGBM, XGBoost, CatBoost, explainable boosting machines (EBM/GAM), points scorecards and MLPs, matched to data size, width, categorical and high-cardinality share, missingness, class imbalance, text, and constraints such as interpretability, monotonic constraints, calibrated probabilities, latency budgets and tiny deployable models. Ranks candidates with starting hyperparameters and cautions, then generates a bake-off script running every candidate on identical CV folds with per-fold metrics, fit time and prediction latency.
clustering-algorithm-selector
Choose the right clustering algorithm and check that clustering makes sense at all: cluster tendency (Hopkins statistic with a correlated-features caveat, nearest-neighbour isolation), then k-means / MiniBatchKMeans, Gaussian mixtures with BIC, agglomerative (Ward, average, complete, single) with dendrograms, DBSCAN, HDBSCAN, OPTICS, spectral clustering, BIRCH, mean shift, k-prototypes and k-modes for mixed or categorical data, and Gower distance, matched to data size, expected shapes, noise, varying density, known k, soft memberships, hierarchy and streaming. Covers preprocessing (scaling, encoding, dimensionality, distance metrics), validation (internal indices, bootstrap stability, external labels, business profiling) and assigning new points.
collaborative-filtering-implementer
Implement collaborative filtering recommenders that beat popularity: item-based kNN with cosine, Jaccard or asymmetric similarity, shrinkage, neighbourhood size, positive-only or mean-centred rating signals and 'because you liked X' explanations; implicit ALS with confidence weighting, BPR pairwise ranking and the closed-form EASE model in plain NumPy, tuned on a temporal validation split by recall@k and NDCG@k and refit on all data; plus user-kNN, explicit matrix factorisation, the implicit/LightFM libraries, hyperparameter ranges, cold-start fallbacks, scaling with sparse matrices and approximate nearest neighbours. Produces top-k lists in the format of the recommendation evaluator.