ch 03 / 12Learning paradigms
slide 01 / 07
Overview

The four families

ParadigmGivenGoalTypical algorithms
Supervised(x, y) pairspredict y from xlinear/logistic regression · trees · SVM · CNN · DETR
Unsupervisedx onlyfind structurek-means · DBSCAN · PCA · GMM · autoencoders
Self-supervisedx only, with pretextlearn representationsMAE · DINO · CLIP · SimCLR · BERT
Reinforcementstates · actions · rewardmaximise long-term rewardQ-learning · policy gradient · PPO · SAC

Most of perception in Black Bee is supervised; the next chapters unpack each paradigm.