41040 - Modular Partition

N. Lygeros

  • Post Category:Articles

A typical deep reinforcement learning agent consists of a large number of interacting components which tend to interact in subtle ways. This is one of the aspects of the difficulty to identify bugs in large computational graphs. One way to solve this problem is to construct common building blocks which are well implemented to get a modular partition of the initial graph. To manage this task it is possible to create a library with mathematical tools.