No major variations have been noticed in the performance of the DLR agent with HLR relative to the random policy for both likelihood and log likelihood reward functions as the number of items was increased.

San Diego State University

1. With EFC student model
2. With HLR student model
3. With DASH (GPL) student model

 Performance of DRL agent when the number of items are varied (Using deep reinforcement learning for personalizing review sessions on e-learning platforms with spaced repetition)

When the number of items was increased, there was not significant increase in  the performances of the DRL agent with EFC relative to the random policy.(likelihood as the reward function).  In case of log likelihood reward function there DLR agent with the EFC performance was slightly decreased relative to the random policy. 

With EFC student model (Using deep reinforcement learning for personalizing review sessions on e-learning platforms with spaced repetition)

With HLR student model (Using deep reinforcement learning for personalizing review sessions on e-learning platforms with spaced repetition)

Similar to the previous student models  no major variations have been noticed in the performance of the DRL agent with the GPL relative to the random policy with  the likelihod and log likelihood reward function.

Learn Before

Related