mlmath — Mathematics for Machine Learning, Interactively

A terminal-based interactive learning tool covering the complete mathematics of machine learning, inspired by Marc Peter Deisenroth's "Mathematics for Machine Learning" (Cambridge University Press, 2020).

---

Features

• 20 curriculum blocks — theory, formulas, derivations, worked examples, and code
• 3-layer visualizations — ASCII (always), plotext terminal plots, matplotlib windows
• Interactive exercises — beginner / intermediate / advanced, with hints and solutions
• Progress tracker — marks blocks as completed, shown as a progress bar
• Topic search — find any block by keyword

---

Quick Start

1. Install dependencies

pip install -r requirements.txt

2. Run

cd mlmath
python main.py

---

Requirements

Package	Purpose
numpy	Core numerical computing
scipy	Special functions, statistics
sympy	Symbolic math, LaTeX rendering
scikit-learn	Reference implementations & data
rich	Terminal UI (tables, markdown)
plotext	Terminal-native plots (Layer 2)
matplotlib	GUI plots in separate window (L3)

---

Curriculum Overview

#	Block	Topics
01	Linear Algebra	Vectors, matrices, norms, SVD, PCA
02	Matrix Decompositions	LU, QR, Cholesky, eigendecomposition
03	Calculus & Autodiff	Gradients, Jacobians, Taylor, dual numbers
04	Optimization	GD, Newton, Lagrange multipliers, convexity
05	Probability Theory	Distributions, Bayes, MLE, conjugate priors
06	Statistics	Estimators, CI, hypothesis tests, p-values
07	Information Theory	Entropy, KL divergence, mutual information
08	Backpropagation	Computational graphs, chain rule, autograd
09	Activation Functions	ReLU, Sigmoid, GELU, Swish, Mish
10	Supervised Learning	Linear/logistic regression, SVM, decision trees
11	Bias-Variance Tradeoff	Decomposition, regularization, cross-validation
12	Unsupervised Learning	K-means, PCA, DBSCAN, autoencoders
13	EM Algorithm	Latent variables, E/M steps, GMM
14	Probabilistic ML	Bayesian inference, GPs, variational inference
15	RL Mathematics	MDPs, Bellman equations, Q-learning, policy gradient
16	MDP Solvers	Value/policy iteration, dynamic programming
17	Deep Learning Math	Attention, batch norm, transformers, residuals
18	NLP Mathematics	TF-IDF, word2vec, BERT, BPE, BLEU
19	Kernel Methods	Mercer's theorem, SVM dual, kernel PCA, GPs
20	Model Evaluation	ROC/AUC, calibration, AIC/BIC, stat tests

---

Project Structure

mlmath/
├── main.py                # Entry point — run this
├── config.py              # Colors, themes, settings
├── requirements.txt
│
├── blocks/                # 20 curriculum blocks
│   ├── __init__.py        # BLOCKS registry
│   ├── b01_linear_algebra.py
│   ├── b02_matrix_decomp.py
│   │   ...
│   └── b20_model_evaluation.py
│
├── exercises/             # Interactive exercises
│   ├── __init__.py        # all_exercises dict
│   ├── ex01_linear_algebra.py
│   ├── ex03_calculus.py
│   ├── ex05_probability.py
│   ├── ex08_backprop.py
│   ├── ex10_supervised.py
│   ├── ex12_unsupervised.py
│   ├── ex15_rl.py
│   └── ex17_deep_learning.py
│
├── core/                  # Math utility library
│   ├── linalg.py
│   ├── calculus.py
│   ├── optimization.py
│   ├── probability.py
│   └── stats.py
│
├── ui/                    # Terminal UI components
│   ├── colors.py
│   ├── widgets.py
│   └── menu.py
│
└── viz/                   # 3-layer visualization
    ├── ascii_plots.py     # Layer 1: always available
    ├── terminal_plots.py  # Layer 2: plotext
    └── matplotlib_plots.py # Layer 3: GUI window

---

Navigation

Main Menu
  [1] Browse Curriculum Blocks  →  select 1-20 → full topic menu → topics
  [2] Practice Exercises        →  pick exercise set → attempt / hint / run
  [3] Search Topics             →  keyword search across all 20 blocks
  [4] Mark Block Complete       →  update progress bar
  [q] Quit

Within each block:

• Topics listed as a numbered menu — pick any topic for theory + example + viz
• [0] goes back to the previous menu

Within each exercise:

• [h] show hint
• [c] show starter code
• [s] reveal full solution
• [r] run solution and print output
• [b] go back

---

Adding a New Block

1. Create blocks/b21_your_topic.py with a run() function:

def run():
    topics = [("Topic Name", topic_fn), ...]
    while True:
        # print menu
        # dispatch by number

2. Register in blocks/__init__.py:

from blocks.b21_your_topic import run as b21
BLOCKS.append(("Your Topic", b21, "Short description"))

3. Add to main.py's BLOCK_META list and block_modules list.

---

Adding a New Exercise Set

1. Create exercises/exXX_topic.py following the Exercise class pattern:

class Exercise:
    def __init__(self, title, difficulty, description, hint, starter_code, solution_code): ...

exercises = [Exercise(...), Exercise(...), Exercise(...)]  # 3 per set

def run(): ...
def _run_exercise(ex): ...

2. Register in exercises/__init__.py:

from exercises.exXX_topic import run as exXX
all_exercises["XX"] = exXX

---

Reference

• Deisenroth, Faisal, Ong — Mathematics for Machine Learning (2020) https://mml-book.github.io/ (free PDF)
• Course website: https://mml-book.com

---

License

MIT License. Educational use encouraged.