Transferring Deep Reinforcement Learning with Adversarial Objective and Augmentation
碩士 === 國立臺灣大學 === 資訊管理學研究所 === 106 === In the past few years, deep reinforcement learning has been proven that can solve problems which have complex states like video games or board games. The next step of intelligent agents would be able to generalize between tasks, using prior experience to pick u...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2018
|
Online Access: | http://ndltd.ncl.edu.tw/handle/xr95cu |
id |
ndltd-TW-106NTU05396048 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-106NTU053960482019-07-25T04:46:48Z http://ndltd.ncl.edu.tw/handle/xr95cu Transferring Deep Reinforcement Learning with Adversarial Objective and Augmentation 利用對抗式目標與資料擴增於深度強化學習間的遷移 Shu-Hsuan Hsu 許書軒 碩士 國立臺灣大學 資訊管理學研究所 106 In the past few years, deep reinforcement learning has been proven that can solve problems which have complex states like video games or board games. The next step of intelligent agents would be able to generalize between tasks, using prior experience to pick up new skills more quickly. However, most reinforcement learning algorithms for now are often suffering from catastrophic forgetting even when facing a very similar target task. Our approach enables the agents to generalize knowledge from a single source task, and boost the learning progress with a semi-supervised learning method when facing a new task. We evaluate this approach on Atari games, a popular reinforcement learning benchmark, and show that it outperforms common baselines based on pre-training and fine-tuning. Bing-Yu Chen 陳炳宇 2018 學位論文 ; thesis 32 zh-TW |
collection |
NDLTD |
language |
zh-TW |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 國立臺灣大學 === 資訊管理學研究所 === 106 === In the past few years, deep reinforcement learning has been proven that can solve problems which have complex states like video games or board games. The next step of intelligent agents would be able to generalize between tasks, using prior experience to pick up new skills more quickly. However, most reinforcement learning algorithms for now are often suffering from catastrophic forgetting even when facing a very similar target task.
Our approach enables the agents to generalize knowledge from a single source task, and boost the learning progress with a semi-supervised learning method when facing a new task.
We evaluate this approach on Atari games, a popular reinforcement learning benchmark, and show that it outperforms common baselines based on pre-training and fine-tuning.
|
author2 |
Bing-Yu Chen |
author_facet |
Bing-Yu Chen Shu-Hsuan Hsu 許書軒 |
author |
Shu-Hsuan Hsu 許書軒 |
spellingShingle |
Shu-Hsuan Hsu 許書軒 Transferring Deep Reinforcement Learning with Adversarial Objective and Augmentation |
author_sort |
Shu-Hsuan Hsu |
title |
Transferring Deep Reinforcement Learning with Adversarial Objective and Augmentation |
title_short |
Transferring Deep Reinforcement Learning with Adversarial Objective and Augmentation |
title_full |
Transferring Deep Reinforcement Learning with Adversarial Objective and Augmentation |
title_fullStr |
Transferring Deep Reinforcement Learning with Adversarial Objective and Augmentation |
title_full_unstemmed |
Transferring Deep Reinforcement Learning with Adversarial Objective and Augmentation |
title_sort |
transferring deep reinforcement learning with adversarial objective and augmentation |
publishDate |
2018 |
url |
http://ndltd.ncl.edu.tw/handle/xr95cu |
work_keys_str_mv |
AT shuhsuanhsu transferringdeepreinforcementlearningwithadversarialobjectiveandaugmentation AT xǔshūxuān transferringdeepreinforcementlearningwithadversarialobjectiveandaugmentation AT shuhsuanhsu lìyòngduìkàngshìmùbiāoyǔzīliàokuòzēngyúshēndùqiánghuàxuéxíjiāndeqiānyí AT xǔshūxuān lìyòngduìkàngshìmùbiāoyǔzīliàokuòzēngyúshēndùqiánghuàxuéxíjiāndeqiānyí |
_version_ |
1719230019485564928 |