Skip to content
LearnMLByDoing
Free
01

Mathematical Foundations

The linear algebra, calculus and probability that every model is built from — learned by bending space with your own matrices.

Lessons
5
Study time
2.7 h
Access
First lesson free

Syllabus

  1. 1.1

    Linear transformations, eigenvectors and SVD in 3D

    Watch a 3×3 matrix shear the unit sphere into an ellipsoid, find the directions it leaves untouched, and factor it into rotate, stretch, rotate.

    35 minFreeLive 3D lab and Python challengeintermediate
    • Read a matrix as a function that moves every point in space
    • Find eigenvectors as the directions a transformation only stretches
    • Decompose any matrix as A = UΣVᵀ and interpret each factor geometrically
  2. 1.2

    Gradients, Jacobians and the chain rule

    Local linear approximations of functions between vector spaces, and why backprop is just the chain rule applied in order.

    30 minProLive 3D lab and Python challengeintermediate
    • Compute gradients of scalar fields and Jacobians of vector fields
    • Compose Jacobians with the multivariate chain rule
    • Verify analytic gradients with finite differences
  3. 1.3

    Probability, Gaussians and maximum likelihood

    Densities as landscapes: covariance ellipsoids, conditioning, and how least squares falls out of a Gaussian likelihood.

    32 minProLive 3D lab and Python challengeintermediate
    • Interpret the covariance matrix as the shape of a Gaussian cloud
    • Derive the MLE of a Gaussian's mean and covariance
    • Connect mean squared error to Gaussian noise assumptions
  4. 1.4

    Convexity, Lagrange multipliers and duality

    When a minimum is guaranteed to be global, and how constraints reshape the landscape — the math behind SVMs and regularisation.

    34 minProLive 3D lab and Python challengeadvanced
    • Recognise convex sets and functions from their Hessians
    • Solve constrained problems with Lagrange multipliers
    • Read L1/L2 regularisation as constrained optimisation
  5. 1.5

    Entropy, cross-entropy and KL divergence

    Surprise, measured in bits: why cross-entropy is the classification loss, what KL divergence costs, and where mutual information shows up.

    30 minProLive 3D lab and Python challengeintermediate
    • Compute entropy, cross-entropy and KL divergence for discrete distributions
    • Show that minimising cross-entropy is maximum likelihood
    • Explain the asymmetry of KL and what forward versus reverse KL prefer