AI Ecosystem CourseBy Queensly Austin
Module 1

How Machine Learning Learns

Introduction Available · Lessons In Development

Start with a question you can test

We have seen that training changes a model and inference uses it. This module examines how the model changes. We will start with a model small enough to calculate by hand, then work toward neural networks, evaluation, and the decisions a business makes with the result. Learning becomes easier to follow when we can see what changes and why.

You will learn to identify a model's inputs, objective, parameters, update procedure, and evaluation. We will also look at misleading performance results and connect prediction errors to operating decisions. No calculus or coding is assumed; new notation comes with words and numerical examples. Google's Machine Learning Crash Course and PyTorch's learning-loop tutorial offer useful practical routes for further study.

Planned lessons

  1. A Model Is A Rule For Producing An Output
  2. Examples, Features, Labels, And Datasets
  3. Loss Tells The Learning Process What To Improve
  4. A Parameter Update You Can Calculate
  5. Batches, Epochs, And The Repeated Learning Loop
  6. Generalization Is The Goal Beyond The Examples
  7. Training, Validation, And Test Sets
  8. Classification And The Confusion Matrix
  9. Thresholds Connect Scores To Actions
  10. Probability, Confidence, And Calibration
  11. Vectors And Matrices Without Assumed Mathematics
  12. Neural Networks And Nonlinear Transformations
  13. Different Learning Setups Solve Different Problems
  14. From Statistical Performance To A Useful Decision