Trendon Learning Atlas

python-for-ml Python for ML Python language and project patterns that keep ML code testable, typed, logged, and production-ready. Future feature idea: Notebook-to-production refactor playground with typing and pytest failures.

shell static shell now Open module

sql-databases SQL and databases Relational querying, indexing, transactions, OLTP/OLAP modeling, and vector search basics. Future feature idea: Query-plan visualizer for joins, windows, indexes, and vector search.

shell backend needed later Open module

data-formats-etl Data formats, ETL, and pipelines CSV, JSONL, Parquet, schema evolution, idempotent pipelines, lineage, backfills, and data quality. Future feature idea: Pipeline simulator showing schema drift, backfills, idempotency, and late data.

shell backend needed later Open module

linear-algebra Linear algebra Vectors, matrices, tensors, projections, cosine similarity, PCA, SVD, rank, and attention math. Future feature idea: Embedding geometry board for cosine similarity, projections, PCA, and low-rank approximation.

shell static shell now Open module

calculus-optimization Calculus and optimization Derivatives, gradients, chain rule, SGD, Adam, saddle points, clipping, and unstable learning rates. Future feature idea: Gradient descent playground with learning rate, momentum, Adam, clipping, and saddle points.

shell static shell now Open module

probability-statistics Probability and statistics Random variables, distributions, Bayes, MLE/MAP, confidence intervals, bootstrap, Monte Carlo, and uncertainty. Future feature idea: Distribution, likelihood, confidence interval, bootstrap, and Monte Carlo explorer.

shell static shell now Open module

eda Exploratory data analysis Summary statistics, distributions, outliers, missing values, duplicates, leakage, imbalance, and train/test mismatch. Future feature idea: Dataset profiler for missing values, outliers, leakage hints, imbalance, and distribution mismatch.

shell backend needed later Open module

data-cleaning Data cleaning Missing-value strategies, outlier handling, clipping, denoising, scaling, log transforms, and rare categories. Future feature idea: Missing-value and outlier strategy lab with before/after metric changes.

shell backend needed later Open module

feature-engineering Feature engineering Categorical, numeric, time, text, interaction, and dimensionality-reduction features without leakage. Future feature idea: Encoding, scaling, target leakage, and dimensionality-reduction sandbox.

shell backend needed later Open module

ml-concepts Core ML concepts Supervised, unsupervised, semi/self-supervised, bias-variance, splits, baselines, metrics, and interpretability. Future feature idea: Bias-variance and train/validation/test split simulator.

shell static shell now Open module

regression Regression Linear, polynomial, Ridge, Lasso, ElasticNet, logistic regression, assumptions, and loss functions. Future feature idea: Regularization playground for L1/L2, train/test error, coefficients, and decision boundary shape.

shell static shell now Open module

classification Classification metrics Confusion matrix, accuracy, precision, recall, F1, ROC-AUC, PR-AUC, thresholds, calibration, and imbalance. Future feature idea: Threshold tuner for precision, recall, F1, ROC-AUC, PR-AUC, and calibration.

shell static shell now Open module

tree-models Tree-based models Decision trees, entropy, Gini, random forests, bagging, boosting, XGBoost, LightGBM, and CatBoost. Future feature idea: Decision tree, random forest, and boosting comparison with feature importance.

shell backend needed later Open module

unsupervised-learning Unsupervised learning K-means, DBSCAN, hierarchical clustering, GMMs, PCA, t-SNE, UMAP, and anomaly detection. Future feature idea: Clustering lab comparing K-means, DBSCAN, PCA, t-SNE, UMAP, and anomaly detection.

shell backend needed later Open module

recommenders-ranking Recommender systems and ranking Collaborative filtering, content-based retrieval, matrix factorization, cold start, ranking, and top-k metrics. Future feature idea: Candidate generation and ranking sandbox with NDCG, MAP@K, and Recall@K.

shell backend needed later Open module

neural-network-foundations Neural network foundations Perceptrons, MLPs, computational graphs, backprop, activations, initialization, optimizers, and regularization. Future feature idea: Forward/backprop visualizer with activations, loss, dropout, initialization, and optimizer choices.

shell static shell now Open module

deep-learning-architectures Deep learning architectures Feedforward nets, autoencoders, VAEs, CNNs, RNNs, Seq2Seq, attention, transformers, GANs, and diffusion. Future feature idea: Architecture map comparing MLP, autoencoder, CNN, RNN, Transformer, GAN, and diffusion.

shell static shell now Open module

cnn CNN Convolution, kernels, stride, padding, pooling, receptive field, channels, ResNet, augmentation, and transfer learning. Future feature idea: Convolution kernel explorer with stride, padding, receptive field, pooling, and transfer learning.

shell static shell now Open module

sequence-models RNN and sequence models Hidden state, BPTT, teacher forcing, LSTM/GRU gates, encoder-decoder models, and attention over states. Future feature idea: Sequence model timeline for hidden state, BPTT, LSTM/GRU gates, and attention over states.

shell static shell now Open module

transformers Transformers Tokens, embeddings, positional encoding, self-attention, Q/K/V, masks, encoders, decoders, generation, and KV cache. Future feature idea: Self-attention playground showing tokens, Q/K/V, masks, heads, positional encoding, and KV cache.

shell static shell now Open module

classical-nlp Classical NLP Tokenization, normalization, n-grams, BoW, TF-IDF, edit distance, Naive Bayes, topic modeling, NER, POS, and parsing. Future feature idea: Text preprocessing and TF-IDF lab with n-grams, edit distance, Naive Bayes, and topic modeling.

shell backend needed later Open module

word-embeddings Word embeddings Distributional hypothesis, Word2Vec, CBOW, Skip-gram, negative sampling, GloVe, FastText, subwords, and OOV. Future feature idea: Word2Vec/FastText embedding space with analogy and out-of-vocabulary examples.

shell backend needed later Open module

deep-nlp Deep NLP RNN language models, Seq2Seq, transformer LMs, MLM/CLM, classification, token tagging, QA, summarization, translation, and retrieval QA. Future feature idea: Task board for language modeling, sequence classification, QA, summarization, translation, and semantic search.

shell backend needed later Open module

tokenization Tokenization for modern LLMs Characters, words, subwords, BPE, WordPiece, Unigram, vocabulary, special tokens, padding, truncation, and masks. Future feature idea: BPE/WordPiece/Unigram tokenizer comparator for Russian and English text plus token cost estimates.

shell backend needed later Open module

llms LLMs, RAG, agents, and evals Pretraining, fine-tuning, SFT, RLHF, DPO, LoRA, quantization, RAG, vector databases, reranking, tool use, agents, safety, and evals. Future feature idea: RAG pipeline workbench with chunking, embeddings, reranking, hallucination tests, and cost/latency tradeoffs.

shell backend needed later Open module

speech-audio-nlp Speech and audio NLP Waveforms, sampling rate, Fourier/STFT, spectrograms, Mel scale, MFCC, CTC, ASR, TTS, diarization, VAD, wav2vec, and Whisper-like models. Future feature idea: Waveform-to-spectrogram explorer with sampling rate, MFCC, CTC alignment, ASR, and TTS concepts.

shell backend needed later Open module

generative-models Diffusion and generative models Autoencoders, VAEs, GANs, normalizing flows, forward noising, reverse denoising, score matching, U-Net, guidance, latent diffusion, and CLIP. Future feature idea: GAN/VAE/diffusion comparison with noising, denoising, guidance, latent diffusion, and CLIP conditioning.

shell backend needed later Open module

ml-lifecycle ML lifecycle Problem framing, data collection, labeling, validation, feature engineering, training, registry, deployment, monitoring, retraining, rollback, and governance. Future feature idea: End-to-end lifecycle board from problem framing and labeling to registry, deployment, monitoring, and retraining.

shell backend needed later Open module

experiment-management Experiment management Reproducibility, seeds, deterministic training, experiment tracking, artifacts, model registry, config management, and tuning. Future feature idea: Experiment tracker demo with seeds, configs, metrics, artifacts, and hyperparameter tuning.

shell backend needed later Open module

data-model-versioning Data and model versioning Code, datasets, features, offline/online consistency, training-serving skew, schema validation, and data contracts. Future feature idea: Version graph for code, data, features, schemas, training-serving skew, and contracts.

shell backend needed later Open module

deployment-patterns Deployment patterns Batch, online, streaming, REST/gRPC, model servers, Docker, Kubernetes, autoscaling, GPU serving, canary, blue-green, shadow, A/B, rollback, warmup, batching, and caching. Future feature idea: Batch vs online vs streaming inference simulator with canary, blue-green, shadow, rollback, caching, and batching.

shell backend needed later Open module

monitoring Monitoring Service metrics, ML metrics, data drift, concept drift, calibration, LLM cost/tokens/latency/tool success/retrieval quality, alerts, dashboards, logs, traces, and OpenTelemetry. Future feature idea: Service, ML, and LLM metric dashboard with drift, calibration, retrieval quality, traces, and alerts.

shell backend needed later Open module

llmops LLMOps Prompt versioning, evals, golden datasets, RAG evals, retrieval metrics, traces, tool logs, failure taxonomies, guardrails, moderation, caching, rate limits, fallbacks, cost control, HITL, and synthetic eval generation. Future feature idea: Prompt/eval/trace/cost dashboard with golden datasets, RAG evals, guardrails, rate limits, and fallbacks.

shell backend needed later Open module

ml-experiments-backlog ML experiment backlog A future server-run backlog for tabular baselines, leakage experiments, imbalance handling, PCA, and calibration. Future feature idea: Backlog runner for baseline models, leakage, imbalance, PCA, calibration, and reproducible results.

shell backend needed later Open module

nlp-experiments-backlog NLP experiment backlog A future backlog for sentiment classification, tokenization comparisons, semantic search, RAG, and fine-tuning comparisons. Future feature idea: Experiment cards for sentiment classification, tokenization, semantic search, RAG, and fine-tuning comparisons.

shell backend needed later Open module

mlops-experiments-backlog MLOps experiment backlog A future backlog for pipeline, drift, monitoring, CI/CD, model gates, Docker, deployment, LLMOps traces, cost, and failure clustering. Future feature idea: Pipeline, monitoring, CI/CD, and LLMOps experiment board with status and artifact links.

shell backend needed later Open module

Trendon Learning Atlas

Data Foundations

Math for ML

Data Preparation

Classical Machine Learning

Deep Learning

NLP

Generative Models

MLOps

Practical AI Agent Layer