Learning
Teaching the robot instead of programming it: data with labels, nearest neighbour, a tiny neural network, learning by reward, models as text.
Start in the simulator- 8.1 See as numbers What a model gets: the sensor views as lists of numbers.
- 8.2 Collecting data Readings with labels, and why bad data makes bad robots.
- 8.3 Nearest neighbour The simplest classifier, in a dozen lines.
- 8.4 A tiny network Weights, a loss, and watching it fall: training in the browser.
- 8.5 Learning by reward Q-learning: a table of how good each action is, filled in by trying.
- 8.6 Tuning itself Learning a controller gain by trial: measure, change, keep the better.
- 8.7 Models as text Saving and loading a model with JSON; models trained elsewhere.
- 8.8 Project: situations A classifier decides what the robot does next.