Salvato in:
Dettagli Bibliografici
Autori principali: Luo, Haozheng, Wu, Tianyi, Han, Colin Feiyu, Yan, Zhijun
Natura: Preprint
Pubblicazione: 2022
Soggetti:
Accesso online:https://arxiv.org/abs/2206.05860
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866929722941767680
author Luo, Haozheng
Wu, Tianyi
Han, Colin Feiyu
Yan, Zhijun
author_facet Luo, Haozheng
Wu, Tianyi
Han, Colin Feiyu
Yan, Zhijun
contents In this work, we build recent advances in distributional reinforcement learning to give a state-of-art distributional variant of the model based on the IQN. We achieve this by using the GAN model's generator and discriminator function with the quantile regression to approximate the full quantile value for the state-action return distribution. We demonstrate improved performance on our baseline dataset - 57 Atari 2600 games in the ALE. Also, we use our algorithm to show the state-of-art training performance of risk-sensitive policies in Atari games with the policy optimization and evaluation.
format Preprint
id arxiv_https___arxiv_org_abs_2206_05860
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle IGN : Implicit Generative Networks
Luo, Haozheng
Wu, Tianyi
Han, Colin Feiyu
Yan, Zhijun
Machine Learning
Artificial Intelligence
In this work, we build recent advances in distributional reinforcement learning to give a state-of-art distributional variant of the model based on the IQN. We achieve this by using the GAN model's generator and discriminator function with the quantile regression to approximate the full quantile value for the state-action return distribution. We demonstrate improved performance on our baseline dataset - 57 Atari 2600 games in the ALE. Also, we use our algorithm to show the state-of-art training performance of risk-sensitive policies in Atari games with the policy optimization and evaluation.
title IGN : Implicit Generative Networks
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2206.05860