Machine Learning

130 topics across 6 chapters

Chapter 1

Math Foundations for ML

Linear Algebra Essentials

3 subtopics

Vectors, matrices, and operations (dot, transpose, inverse intuition)

Eigenvalues/eigenvectors and SVD intuition (why they matter in ML)

Matrix calculus basics (gradients for linear models)

Calculus & Optimization Basics

3 subtopics

Derivatives, partial derivatives, and the chain rule

Gradient descent and learning rate behavior

Convexity intuition and common loss functions

Probability & Statistics for ML

4 subtopics

Random variables and common distributions (Gaussian, Bernoulli, Poisson)

Expectation, variance, covariance, correlation

Bayes’ rule and conditional probability (posterior intuition)

Statistical estimation (MLE/MAP) and confidence vs. credible intervals

Numerical Computing Skills

3 subtopics

Floating point, stability, and conditioning (why training can diverge)

Vectorization mindset (broadcasting and avoiding Python loops)

Automatic differentiation basics (what backprop is computing)

Chapter 2

Core ML Concepts & Workflow

Problem framing and success metrics

3 subtopics

Choose task type (classification, regression, ranking, forecasting)

Define baseline and target metric (accuracy, F1, AUROC, RMSE, etc.)

Set constraints (latency, memory, interpretability, cost)

Data splitting and evaluation hygiene

3 subtopics

Train/validation/test splits and leakage patterns

Cross-validation (when to use it and pitfalls)

Calibration and thresholding for decision-making

Generalization, bias–variance, and regularization

3 subtopics

Underfitting vs overfitting (diagnosis with learning curves)

Regularization methods (L1/L2, early stopping, dropout intuition)

Hyperparameter tuning basics (search spaces and budgets)

Feature engineering and preprocessing

4 subtopics

Scaling/normalization and handling missing values

Encoding categorical variables (one-hot, target, embeddings overview)

Text/vector representations overview (bag-of-words to embeddings)

Feature selection and dimensionality reduction basics

Metrics and error analysis

3 subtopics

Confusion matrix, precision/recall, ROC/PR curves

Residual analysis for regression and heteroscedasticity clues

Slice-based evaluation (subgroups, rare cases, long tail)

Reproducibility and experiment tracking

2 subtopics

Random seeds, deterministic ops, and versioning data/code

Run logging (configs, metrics) and comparing experiments

Chapter 3

Supervised Learning

Linear Models

3 subtopics

Linear regression (loss, closed form vs GD)

Logistic regression (decision boundary, calibration)

Regularized models (Ridge/Lasso/Elastic Net)

Tree-Based Methods

3 subtopics

Decision trees (splits, impurity, overfitting controls)

Random forests (bagging, OOB error, feature importance caveats)

Gradient boosting (XGBoost/LightGBM/CatBoost concepts)

Kernel Methods and Margin-Based Models

2 subtopics

SVM basics (margin, kernels, C and gamma intuition)

Kernel ridge regression and Gaussian processes (high-level)

Nearest Neighbors and Similarity

2 subtopics

kNN (distance metrics, scaling sensitivity)

Metric learning intuition (why embeddings help)

Model Interpretation for Supervised Learning

3 subtopics

Permutation importance and partial dependence (when they mislead)

SHAP/LIME overview and practical pitfalls

Interpreting linear/logistic models (coefficients, odds ratios)

Chapter 4

Unsupervised & Self-Supervised Learning

Clustering

3 subtopics

k-means (initialization, scaling, choosing k)

Hierarchical clustering and dendrogram interpretation

Density-based methods (DBSCAN/HDBSCAN intuition)

Dimensionality Reduction

3 subtopics

PCA (variance, SVD connection)

t-SNE/UMAP (visualization-focused caveats)

Autoencoders basics (representation learning)

Representation Learning (Self-Supervised)

2 subtopics

Contrastive learning intuition (positives/negatives, collapse)

Pretext tasks and masked prediction (high-level)

Generative Modeling Basics

3 subtopics

Likelihood-based models (autoregressive overview)

VAEs (encoder/decoder, KL term intuition)

Diffusion models intuition (noise-to-data idea)

Chapter 5

Deep Learning

Neural Network Fundamentals

4 subtopics

Perceptron, MLPs, activations (ReLU, sigmoid, GELU)

Forward/backprop and computation graphs (conceptual + simple derivation)

Initialization, normalization (BatchNorm/LayerNorm) basics

Optimizers (SGD+momentum, Adam) and training stability

Convolutional Neural Networks (Vision)

3 subtopics

Convolutions, padding/stride, receptive fields

Modern CNN blocks (residual connections, depthwise separable conv)

Data augmentation and transfer learning for vision

Sequence Models

3 subtopics

RNN/LSTM/GRU intuition (vanishing gradients)

Attention mechanism basics

Transformers overview (encoder/decoder, positional encoding)

Training Practice & Debugging

4 subtopics

Overfit a small batch and sanity checks

Diagnose gradient issues (vanishing/exploding, clipping)

Data pipeline bugs (shapes, labels, normalization mismatches)

Compute planning (batch size, mixed precision, throughput)

Deep Learning for Text (NLP)

3 subtopics

Tokenization (BPE/WordPiece overview) and embeddings

Fine-tuning vs prompt-based usage (high-level tradeoffs)

Evaluation for NLP (BLEU/ROUGE vs task-specific + human eval)

Generative Deep Learning

3 subtopics

↗ VAEs (encoder/decoder, KL term intuition) (see Chapter 4)

↗ Diffusion models intuition (noise-to-data idea) (see Chapter 4)

GANs (generator/discriminator game, mode collapse)

Chapter 6

ML Engineering, Deployment, and Responsible AI

Data & Pipelines

3 subtopics

100

Dataset creation/labeling strategy and quality checks

101

Data versioning and lineage (schemas, snapshots)

102

Feature stores concept (offline/online consistency)

103

Serving and Deployment

3 subtopics

104

Batch vs online inference; latency/throughput tradeoffs

105

Model packaging and inference optimization (ONNX, quantization overview)

106

CI/CD for ML (testing data, models, and pipelines)

107

Monitoring and Maintenance

3 subtopics

108

Data drift and concept drift (detection signals)

109

Model performance monitoring and alerting

110

Retraining strategies and rollout (shadow, canary, A/B tests)

111

Scalable Training and Systems

3 subtopics

112

GPU basics and data loading bottlenecks

113

Distributed training concepts (DDP, data/model parallel)

114

Checkpointing and fault tolerance

115

Responsible AI and Safety

4 subtopics

116

Fairness basics (bias sources, group metrics)

117

Privacy and security (PII handling, membership inference intuition)

118

Robustness to distribution shift and adversarial examples (overview)

119

Model cards and documentation (communicating limitations)

120

Productization and Stakeholders

3 subtopics

121

Translating business goals to ML objectives and KPIs

122

Human-in-the-loop systems (labeling feedback, review queues)

123

Communicating results (uncertainty, demos, failure modes)