ML Foundations Lab

The absolute basics: what a function is, how to read a graph, and the three function families (linear, polynomial, exponential/logarithmic) that show up everywhere in ML.

• ▷ What Is a Function?
• 🔒 Linear Functions: y = mx + b
• 🔒 Polynomials, Exponentials & Logarithms

Vectors and matrices as geometric objects and transformations — the language every ML model is written in.

• 🔒 Vectors: Geometric & Numeric Views
• 🔒 Matrices as Transformations
• 🔒 Multiplication, Identity, Inverse & Determinant
• 🔒 Eigenvectors & Eigenvalues

Slow, deep coverage of the derivative — slope, rate of change, the differentiation rules, and gradients in many dimensions.

• 🔒 Slope of a Curve & Rate of Change
• 🔒 Power, Product & Chain Rules
• 🔒 Partial Derivatives
• 🔒 The Gradient as Steepest Ascent

Area under a curve, the fundamental theorem of calculus, and where integration shows up in probability and expectation.

• 🔒 Area Under a Curve: Riemann Sums
• 🔒 The Fundamental Theorem of Calculus
• 🔒 Integration in Machine Learning

Distributions, expectation, variance, Bayes' theorem, and the intuition behind maximum likelihood.

• 🔒 Distributions: Uniform, Normal & Bernoulli
• 🔒 Expectation, Variance & Bayes' Theorem
• 🔒 Maximum Likelihood Intuition

Linear regression, loss functions, gradient descent, and the overfitting/underfitting tradeoff — the core loop of supervised learning.

• 🔒 Linear Regression
• 🔒 Loss Functions: MSE & MAE
• 🔒 Gradient Descent
• 🔒 Overfitting & Underfitting

Sigmoid, tanh, ReLU and softmax; logistic regression's decision boundary; and why non-linearity is what makes deep networks powerful.

• 🔒 Activation Functions
• 🔒 Logistic Regression
• 🔒 Why Non-Linearity Matters

From a single perceptron to a trainable multilayer network — build one from scratch in the browser and watch backpropagation work, number by number.

• 🔒 Perceptron to Multilayer Network
• 🔒 Train a Network in the Browser
• 🔒 Backpropagation Walkthrough