TY - GEN
T1 - Robust population coding in free-energy-based reinforcement learning
AU - Otsuka, Makoto
AU - Yoshimoto, Junichiro
AU - Doya, Kenji
PY - 2008
Y1 - 2008
N2 - We investigate the properties of free-energy-based reinforcement learning using a new experimental platform called the digit floor task. The simulation results showed the robustness of the reinforcement learning method against noise applied in both the training and testing phases. In addition, reward-dependent and reward-invariant representations were found in the distributed activation patterns of hidden units. The representations coded in a distributed fashion persisted even when the number of hidden nodes were varied.
AB - We investigate the properties of free-energy-based reinforcement learning using a new experimental platform called the digit floor task. The simulation results showed the robustness of the reinforcement learning method against noise applied in both the training and testing phases. In addition, reward-dependent and reward-invariant representations were found in the distributed activation patterns of hidden units. The representations coded in a distributed fashion persisted even when the number of hidden nodes were varied.
UR - http://www.scopus.com/inward/record.url?scp=58849087267&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=58849087267&partnerID=8YFLogxK
U2 - 10.1007/978-3-540-87536-9_39
DO - 10.1007/978-3-540-87536-9_39
M3 - Conference contribution
AN - SCOPUS:58849087267
SN - 3540875352
SN - 9783540875352
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 377
EP - 386
BT - Artificial Neural Networks - ICANN 2008 - 18th International Conference, Proceedings
T2 - 18th International Conference on Artificial Neural Networks, ICANN 2008
Y2 - 3 September 2008 through 6 September 2008
ER -