This project analyzes multimodal data in rodents while animals conducted the social discriminant task. In this task, a subject mouse discriminated a reward-associated mouse from a no-reward-mouse and licked the pot only when the reward-associated mouse was present (Fig. 3A). But when subject mice was administered by a dopamine blocker, their licking behavior in response to the reward-associated mouse disappeared. The project uses several classifiers, including SVM, FDA, Gaussian process and Hidden Markov Model to predict licking behavior and shows that the trained model failed to predict licking in drug-administered mice. It also examines neural data obtained by fiber photometry and classifies them into reward information. The project shows a drop of classification accuracy when learning is reversed (Fig. 3B).