Exploring different reinforcment learning and hueristics for learning a changing space.
Problems can be divided into those that humans can do better than machines and those that machines can do better than humans. Furthermore, there are problems for which the optimal solution is or can be known, and those for which it can't be known.
Dificulties in the interface or hard to define goals:
- driving in a city
- picking fruit
- flipping burgers
With either computable solutions or fewer limitations:
- play go
- fly a plane
- detect cancer in a a scan
Examples in the real world include:
- Fraud and anti money laundering (my background)
- Virus and malware pattern detection