notesum.ai

Published at December 6

A Temporally Correlated Latent Exploration for Reinforcement Learning

cs.LG

cs.AI

Released Date: December 6, 2024

Authors: SuMin Oh¹, WanSoo Kim¹, HyunJin Kim¹

Aff.: ¹School of EEE, Dankook University, Republic of Korea

Arxiv: http://arxiv.org/pdf/2412.04775v1

Refer to caption

With Noisy TV						Without Noisy TV
Environment	-1.0	0.0	0.5	1.0	2.0	Environment	-1.0	0.0	0.5	1.0	2.0
DoorKey $8\times 8$	.697	.379	.318	.536	.565	DoorKey $8\times 8$	.647	.839	.713	.689	.771
DoorKey $16\times 16$	.311	.048	.040	.033	.200	DoorKey $16\times 16$	.209	.041	.294	.019	.286
LCS9N3^[1]	.921	.930	.934	.929	.932	LCS9N3^[1]	.941	.941	.941	.939	.940
LCS11N5^[1]	.000	.000	.000	.000	.000	LCS11N5^[1]	.485	.000	.000	.719	.430
DO $8\times 8$ ^[2]	.536	.929	.884	.903	.947	DO $8\times 8$ ^[2]	.730	.300	.691	.970	.877
DO $16\times 16$ ^[2]	.631	.959	.954	.978	.968	DO $16\times 16$ ^[2]	.819	.807	.958	.956	.897
Empty $8\times 8$	.939	.936	.938	.938	.938	Empty $8\times 8$	.935	.939	.935	.933	.937
Empty $16\times 16$	.921	.913	.901	.912	.927	Empty $16\times 16$	.936	.874	.905	.903	.924
KeyCorridorS3R3	.000	.000	.001	.000	.000	KeyCorridorS3R3	.079	.524	.000	.156	.087
MultiRoomN2S4	.814	.813	.815	.813	.814	MultiRoomN2S4	.827	.827	.828	.828	.823
BankHeist^[3]	.719	.687	.651	.676	.580	SpaceInvaders	.420	.650	.599	.519	.619