| bibtype |
J -
Journal Article
|
| ARLID |
0602819 |
| utime |
20241217090339.0 |
| mtime |
20241216235959.9 |
| SCOPUS |
85210284955 |
| DOI |
10.1109/ACCESS.2024.3497589 |
| title
(primary) (eng) |
Knowledge Transfer in Deep Reinforcement Learning via an RL-Specific GAN-Based Correspondence Function |
| specification |
| page_count |
15 s. |
| media_type |
E |
|
| serial |
| ARLID |
cav_un_epca*0461036 |
| ISSN |
2169-3536 |
| title
|
IEEE Access |
| volume_id |
12 |
| volume |
1 (2024) |
| page_num |
177204-177218 |
| publisher |
| name |
Institute of Electrical and Electronics Engineers |
|
|
| keyword |
Deep learning |
| keyword |
Markov decision process |
| keyword |
reinforcement learning |
| keyword |
transfer learning |
| keyword |
knowledge transfer |
| author
(primary) |
| ARLID |
cav_un_auth*0333672 |
| name1 |
Ruman |
| name2 |
Marko |
| institution |
UTIA-B |
| department |
AS |
| full_dept (cz) |
Adaptivní systémy |
| full_dept (eng) |
Department of Adaptive Systems |
| department (cz) |
AS |
| department (eng) |
AS |
| country |
SK |
| garant |
K |
| fullinstit |
Ústav teorie informace a automatizace AV ČR, v. v. i. |
|
| author
|
| ARLID |
cav_un_auth*0101092 |
| name1 |
Guy |
| name2 |
Tatiana Valentine |
| institution |
UTIA-B |
| full_dept (cz) |
Adaptivní systémy |
| full_dept |
Department of Adaptive Systems |
| department (cz) |
AS |
| department |
AS |
| full_dept |
Department of Adaptive Systems |
| fullinstit |
Ústav teorie informace a automatizace AV ČR, v. v. i. |
|
| source |
|
| cas_special |
| project |
| project_id |
CA21169 |
| agency |
EU-COST |
| country |
XE |
| ARLID |
cav_un_auth*0452289 |
|
| abstract
(eng) |
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. |
| result_subspec |
WOS |
| RIV |
IN |
| FORD0 |
10000 |
| FORD1 |
10200 |
| FORD2 |
10201 |
| reportyear |
2025 |
| num_of_auth |
2 |
| inst_support |
RVO:67985556 |
| permalink |
https://hdl.handle.net/11104/0360153 |
| cooperation |
| ARLID |
cav_un_auth*0478849 |
| name |
Provozně ekonomická fakulta, Česká zemědělská univerzita v Praze |
| institution |
PEF CZU |
| country |
CZ |
|
| confidential |
S |
| mrcbC91 |
A |
| mrcbT16-e |
COMPUTERSCIENCEINFORMATIONSYSTEMS|ENGINEERINGELECTRICALELECTRONIC|TELECOMMUNICATIONS |
| mrcbT16-j |
0.698 |
| mrcbT16-D |
Q3 |
| arlyear |
2024 |
| mrcbU14 |
85210284955 SCOPUS |
| mrcbU24 |
PUBMED |
| mrcbU34 |
WOS |
| mrcbU63 |
cav_un_epca*0461036 IEEE Access 12 1 2024 177204 177218 2169-3536 2169-3536 Institute of Electrical and Electronics Engineers |
|