Basics of ML

Math, Probability & Statistics for ML — Build the Foundation That Makes Everything Else Click

Most people struggle with ML not because the algorithms are hard but because the foundation underneath them was never properly built. Gradient descent makes no sense without calculus. Naive Bayes makes no sense without probability. A/B testing makes no sense without hypothesis testing. This resource builds that foundation — clearly, deliberately, and directly connected to how each concept shows up in real ML work.

What's covered:

Mathematics for ML

1. Linear algebra — vectors, matrices, transformations, eigenvalues and why they matter in PCA and neural networks
2. Calculus — derivatives, partial derivatives, and the chain rule that powers every backpropagation step
3. Optimisation — gradient descent, loss functions, and the math behind how models actually learn
4. Matrix operations — the mechanics behind every prediction your model makes

Probability

1. Foundations — events, sample space, conditional probability, independence
2. Bayes theorem — the single most important formula in probabilistic ML, explained with intuition
3. Probability distributions — Normal, Binomial, Poisson, and where each one appears in algorithms
4. Expectation, variance, and how uncertainty is quantified inside models

Statistics

1. Descriptive statistics — mean, median, mode, variance, standard deviation, skewness
2. Inferential statistics — sampling distributions, standard error, confidence intervals
3. Correlation and covariance — measuring relationships without confusing them with causation
4. How statistical thinking directly shapes feature selection, model evaluation, and result interpretation

Hypothesis Testing

1. Null and alternative hypotheses — structuring a testable question correctly
2. p-values and significance levels — what they actually mean and how they are routinely misread
3. Type I and Type II errors — the two ways a confident conclusion can still be wrong
4. t-tests, z-tests, chi-square tests — choosing the right test for the right situation
5. Confidence intervals — expressing uncertainty honestly instead of hiding it behind a point estimate
6. A/B testing from start to finish — designing the experiment, running the test, reading the result

Who this is for:

Anyone in data, analytics, or ML who wants to stop treating the math as a black box — whether you are just starting your ML journey, preparing for technical interviews, or trying to understand why your model behaves the way it does.

You don't need to be a mathematician. You need to understand enough to build with confidence. This is exactly that.