DSpace Repository

Browsing by Author "Uchibe, Eiji"

Browsing by Author "Uchibe, Eiji"

Sort by: Order: Results:

  • Morimura, Tetsuro; Uchibe, Eiji; Yoshimoto, Junichiro; Doya, Kenji (Nara Institute of Science and Technology奈良先端科学技術大学院大学, 2007-09)
    Conventional policy gradient reinforcement learning (PGRL) algorithms neglect a term in the average reward gradient, which is dependent upon the change in the stationary distribution by the policy parameter change. Although ...