In concept, machine learning = find a function by itself
Classified based on the output of the function
- Regression - value
- Classification - category
- Structured learning / generative learning - structured stuff (image, sentence, etc.)
ChatGPT: decompose a generative learning problem into multiple classification problems
Three steps of machine learning
- Function set / model: deep learning (CNN, Transformer, ...), Decision Tree, etc.
- Goodness of functionn: supervised learning, semi-supervised learning, reinforcement learning, etc.
- Find the best function: gradient descent (Adam, AdamW, ...), genetic algorithm, etc.