Adversarial search

Many games can be fully solved through search techniques but the problem formulation and the search approaches that we have studied so far cannot handle searching in a strategic situation with multiple agents.  In this module, we will discuss adversarial search techniques, starting from simple games that can be fully searched and moving to more advanced games where we must use approximations to the full search.

Hybrid machine learning and search approaches have been in the news lately especially with the AlphaZero series of games (see link and photo to the right).  Although we will discuss the foundational search approaches for these methods, we will not cover the machine learning methods until a later module.

Real World search

Photo from this DeepMind blog post about AlphaZero