- Is the situation stochastic or deterministic?
- Is it time-inhomogeneous? (Different across time?)
- How much data do you have available?
- What limitations are there with respect to computational cost (compute and time), both for training and predicting?
- Do you need to try actions to learn about situations? (If so, consider Reinforcement Learning.)
- Do your actions have an impact on the environment? (If not, Reinforcement Learning may be a poor choice because you are just implicitly learning to predict future variables using weaker reinforcement signals compared to the supervised learning case.)
References: Bayes-Ian, cesarsalgado on Reddit
Context: When doing research for my Machine Learning in Trading project, I came across a Reddit thread on applying (A) DeepMind’s deep learning methods for training a computer to play Atari to (B) the problem of predicting stock prices. This thread contained comments on why that set of reinforcement learning methods would be a bad fit for predicting stock prices. I generalised those comments and put them in the list above.