Preporato
CoursesCertificationsLabs
Preporato
CoursesCertificationsLabs
ML practiceStep 4 of 12

Derivatives and gradients

How a change in one number moves the loss: the chain rule, in code.

Start 5 problems

Problems

  1. 1Numerical gradient by central differencesEasy · 10 min · 11 hidden testsSolution free
  2. 2Sigmoid that never overflowsEasy · 10 min · 10 hidden tests
  3. 3Gradient of mean squared errorMedium · 14 min · 10 hidden tests
  4. 4Softmax cross-entropy gradient for a batchMedium · 15 min · 10 hidden tests
  5. 5Gradient check for a backward passMedium · 12 min · 12 hidden tests

Practice it at full scale

Hands-on labs on real tooling, in your browser.

A PyTorch Training Loop From Scratch
Probability and statisticsGradient descent