Login / Signup
Expert-based reward function training: the novel method to train sequence generators.
Joji Toyama
Yusuke Iwasawa
Kotaro Nakayama
Yutaka Matsuo
Published in:
ICLR (Workshop) (2018)
Keyphrases
</>
dynamic programming
missing values
objective function
pairwise
significant improvement
state space
missing data
convergence rate