Concept

Usefulness of the Average-of-Sum-of-Outcomes Reward Function (Using deep reinforcement learning for personalizing review sessions on e-learning platforms with spaced repetition)

The likelihood-based reward function depends on the student's modeled state, so it cannot be applied unless the student is modeled first. The average-of-sum-of-outcomes reward function avoids this dependency because it does not require the student state, making it usable without prior student modeling. Its drawback is that it becomes computationally expensive as the number of exercises grows; sampling exercises into categories can help mitigate this cost.

0

1

Updated 2026-07-20

Tags

Data Science