Show Thoughts

Weisswange et al. apply the idea of Bayesian inference to multi-modal integration and action selection. They show that online reinforcement learning can effectively train a neural network to approximate a Q-function predicting the reward in a multi-modal cue integration task.