Knowledge Transfer in Deep Reinforcement Learning via an RL-Specific GAN-Based Correspondence Function
Deep reinforcement learning has demonstrated superhuman performance in complex decision-making tasks, but it struggles with generalization and knowledge reuse—key aspects of true intelligence. This article introduces a novel approach that modifies Cycle Generative Adversarial Networks spe...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10752398/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1839631311660122112 |
|---|---|
| author | Marko Ruman Tatiana V. Guy |
| author_facet | Marko Ruman Tatiana V. Guy |
| author_sort | Marko Ruman |
| collection | DOAJ |
| description | Deep reinforcement learning has demonstrated superhuman performance in complex decision-making tasks, but it struggles with generalization and knowledge reuse—key aspects of true intelligence. This article introduces a novel approach that modifies Cycle Generative Adversarial Networks specifically for reinforcement learning, enabling effective one-to-one knowledge transfer between two tasks. Our method enhances the loss function with two new components: model loss, which captures dynamic relationships between source and target tasks, and Q-loss, which identifies states significantly influencing the target decision policy. Tested on the 2-D Atari game Pong, our method achieved 100% knowledge transfer in identical tasks and either 100% knowledge transfer or a 30% reduction in training time for a rotated task, depending on the network architecture. In contrast, using standard Generative Adversarial Networks or Cycle Generative Adversarial Networks led to worse performance than training from scratch in the majority of cases. The results demonstrate that the proposed method ensured enhanced knowledge generalization in deep reinforcement learning. |
| format | Article |
| id | doaj-art-5fe6f68baf904dba9842fefd8edf0efd |
| institution | Matheson Library |
| issn | 2169-3536 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-5fe6f68baf904dba9842fefd8edf0efd2025-07-11T23:00:53ZengIEEEIEEE Access2169-35362024-01-011217720417721810.1109/ACCESS.2024.349758910752398Knowledge Transfer in Deep Reinforcement Learning via an RL-Specific GAN-Based Correspondence FunctionMarko Ruman0https://orcid.org/0000-0001-9349-377XTatiana V. Guy1https://orcid.org/0000-0003-1017-0727Department of Adaptive Systems, Institute of Information Theory and Automation, Czech Academy of Sciences, Prague, Czech RepublicDepartment of Adaptive Systems, Institute of Information Theory and Automation, Czech Academy of Sciences, Prague, Czech RepublicDeep reinforcement learning has demonstrated superhuman performance in complex decision-making tasks, but it struggles with generalization and knowledge reuse—key aspects of true intelligence. This article introduces a novel approach that modifies Cycle Generative Adversarial Networks specifically for reinforcement learning, enabling effective one-to-one knowledge transfer between two tasks. Our method enhances the loss function with two new components: model loss, which captures dynamic relationships between source and target tasks, and Q-loss, which identifies states significantly influencing the target decision policy. Tested on the 2-D Atari game Pong, our method achieved 100% knowledge transfer in identical tasks and either 100% knowledge transfer or a 30% reduction in training time for a rotated task, depending on the network architecture. In contrast, using standard Generative Adversarial Networks or Cycle Generative Adversarial Networks led to worse performance than training from scratch in the majority of cases. The results demonstrate that the proposed method ensured enhanced knowledge generalization in deep reinforcement learning.https://ieeexplore.ieee.org/document/10752398/Deep learningMarkov decision processreinforcement learningtransfer learningknowledge transfer |
| spellingShingle | Marko Ruman Tatiana V. Guy Knowledge Transfer in Deep Reinforcement Learning via an RL-Specific GAN-Based Correspondence Function IEEE Access Deep learning Markov decision process reinforcement learning transfer learning knowledge transfer |
| title | Knowledge Transfer in Deep Reinforcement Learning via an RL-Specific GAN-Based Correspondence Function |
| title_full | Knowledge Transfer in Deep Reinforcement Learning via an RL-Specific GAN-Based Correspondence Function |
| title_fullStr | Knowledge Transfer in Deep Reinforcement Learning via an RL-Specific GAN-Based Correspondence Function |
| title_full_unstemmed | Knowledge Transfer in Deep Reinforcement Learning via an RL-Specific GAN-Based Correspondence Function |
| title_short | Knowledge Transfer in Deep Reinforcement Learning via an RL-Specific GAN-Based Correspondence Function |
| title_sort | knowledge transfer in deep reinforcement learning via an rl specific gan based correspondence function |
| topic | Deep learning Markov decision process reinforcement learning transfer learning knowledge transfer |
| url | https://ieeexplore.ieee.org/document/10752398/ |
| work_keys_str_mv | AT markoruman knowledgetransferindeepreinforcementlearningviaanrlspecificganbasedcorrespondencefunction AT tatianavguy knowledgetransferindeepreinforcementlearningviaanrlspecificganbasedcorrespondencefunction |