Phase 237 DaysIntermediate
Phase 2 — Math & Statistics for ML
Build the mathematical intuition behind every ML algorithm — linear algebra, calculus, probability, and statistics — so you can diagnose and fix models confidently.
- Implement linear algebra operations using NumPy without relying on black-box calls.
- Derive gradient descent from first principles and visualize convergence.
- Apply statistical reasoning to EDA, hypothesis testing, and feature analysis.
⚡ Must Know
- Vectors, Matrices, Tensors — shapes and operations
- Dot Product + Matrix Multiplication
- L1/L2 Norms + Cosine Similarity
- Derivatives + Partial Derivatives
- Chain Rule — foundation of backprop
- Gradient Descent — intuition + implementation
- Mean, Median, Mode, Variance, Std Dev
- Probability Basics + Bayes Theorem
- Normal, Binomial, Uniform Distributions
- Correlation vs Causation
- Hypothesis Testing + p-values
- NumPy — broadcasting, linalg
- Pandas — EDA, groupby, cleaning
- PCA — intuition + sklearn
✨ Good to Know
- Eigenvalues + Eigenvectors
- Entropy + Cross-Entropy
- Hessian + Jacobian (conceptual)
- Monte Carlo Methods
- SVD — Singular Value Decomposition
📚 Resources
3Blue1Brown — Linear Algebra
Best visual intuition for vectors, matrices, and transformations.
youtube.com/3b1b ↗StatQuest with Josh Starmer
ML math explained with simple visuals and clear intuition.
statquest.org ↗🏗️ Projects
Gradient Descent Visualizer
Implement from scratch and animate loss surface convergence across learning rates.
PCA on MNIST
Reduce MNIST dimensionality and measure compression vs accuracy tradeoff.
Bayesian A/B Tester
Bayesian engine estimating probability of lift and decision confidence.