Draft: What Humans Can Learn From Machine Learning
Human Learning differs from Machine Learning in important ways.
However, there are useful things we can take from ML
- the alpha parameter
- momentum: https://en.wikipedia.org/wiki/Stochastic_gradient_descent#Momentum
- optimality in the face of an impossibly large search space
- generalize > memorize by using a testing set
- epsilon exhaustion and Probably Approximately Correct learning
- naive bayes - how can it be right
- So you just need to be directionally correct.
- random restart hill climbing
- if theres a v specific point that is superbly better than others, you're in a bad world
- simulated annealing - hot and cold
- eager learning vs lazy learning
⚠️ You are reading an unpublished, incomplete draft. Questions are welcome but feedback/criticism may be premature.