Role

Software Engineer

Impacts

Designed a reinforcement learning algorithm with custom weighting factors

Skills

Python

Software programming

Reinforcement learning

Algorithms

Data structures

Descriptions

We implemented a Backgammon gameplay agent using reinforcement learning. The backbone algorithm of the agent is expectiminimax search with alpha-beta pruning to maintain its effeciency. Within the algorithm, we incorporated several factors in evaluating the state of the boards. For example, a racing strategy was implemented to check how far away the checkers are from their respective homebase. A few more strategies (bearoff, hitting, stacking, and making primes) were also used in our reinforcement decision factors.