| Title | : | Cost-Sensitive Trees for Interpretable Reinforcement Learning |
| Speaker | : | Nishtala Siddharth (IITM) |
| Details | : | Wed, 12 Jul, 2023 12:00 PM @ SSB 233 (MR-1) |
| Abstract: | : | Trees have emerged as the most popular choice of intrinsically interpretable models to represent policies in reinforcement learning. However, directly learning a tree policy poses challeng es, prompting existing approaches to employ neural network policies to generate datasets for training tree-based models in a supervised manner. Nonetheless, these approaches treat all misclassifications equally, assuming that there is one optimal action while considering all other actions equally sub-optimal. This work presents a novel perspective by associating different costs with various misclassifications. By adopting a cost-sensitive approach to tree construction, we demonstrate that policies generated using this methodology exhibit improved performance. To validate our findings, we develop cost-sensitive variants of two established methods, VIPER and MoET, and provide empirical evidence showcasing their superiority over the original methods across diverse environments. |
