The AI Algorithms Compendium

From uninformed search to diffusion models and large language models, The AI Algorithms Compendium derives, implements, and tests the algorithms that define artificial intelligence today. Written for students, engineers, and self-learners who want more than a high-level overview, this 42-chapter reference pairs a rigorous theoretical treatment (theorems, proofs, and derivations where they matter) with complete, runnable Python code for every method it covers.

The book moves through six parts: classical search, logic, and planning (Traditional AI); Bayesian inference and statistical machine learning (regression, trees, SVMs, clustering, dimensionality reduction); deep learning (CNNs, RNNs, and the Transformer architecture from first principles); reinforcement learning (MDPs, Q-learning, policy gradients, bandits); generative AI (autoencoders, GANs, diffusion models, large language models, fine-tuning, RAG, and agents); and cross-cutting topics including interpretability, AI safety, MLOps, and ethics.

Every chapter follows the same structure: motivation and intuition, the underlying mathematics, a complete working implementation, a worked example, and exercises. All examples use public-domain and openly licensed data (MNIST, Fashion-MNIST, the UCI Machine Learning Repository, Project Gutenberg texts) or synthetic data, so everything can be run and modified freely.

Read it cover to cover, or use it as a lookup reference for a specific algorithm.

Material portions of this book were generated using generative AI tools, then reviewed, edited, designed, and curated by Choral Software.

Companion source code: Download the algorithms and pseudocode from https://pub-e945d7c8d1d640c7be2864f060e0f048.r2.dev/ai-algorithms-compendium-companion-v1.zip