تقرير
Self-Consistent Models and Values
العنوان: | Self-Consistent Models and Values |
---|---|
المؤلفون: | Farquhar, Gregory, Baumli, Kate, Marinho, Zita, Filos, Angelos, Hessel, Matteo, van Hasselt, Hado, Silver, David |
سنة النشر: | 2021 |
المجموعة: | Computer Science Statistics |
مصطلحات موضوعية: | Computer Science - Machine Learning, Computer Science - Artificial Intelligence, Statistics - Machine Learning |
الوصف: | Learned models of the environment provide reinforcement learning (RL) agents with flexible ways of making predictions about the environment. In particular, models enable planning, i.e. using more computation to improve value functions or policies, without requiring additional environment interactions. In this work, we investigate a way of augmenting model-based RL, by additionally encouraging a learned model and value function to be jointly \emph{self-consistent}. Our approach differs from classic planning methods such as Dyna, which only update values to be consistent with the model. We propose multiple self-consistency updates, evaluate these in both tabular and function approximation settings, and find that, with appropriate choices, self-consistency helps both policy evaluation and control. Comment: NeurIPS 2021 |
نوع الوثيقة: | Working Paper |
URL الوصول: | http://arxiv.org/abs/2110.12840 |
رقم الأكسشن: | edsarx.2110.12840 |
قاعدة البيانات: | arXiv |
الوصف غير متاح. |