Application Research of Few-Shot Learning in AI Personalized Educational Resource Recommendation

Now that online education is developing faster and faster, we have found that in order to make learning better, personalized education resource recommendation systems are particularly important. However, the common recommendation algorithms used in education in the past will encounter a lot of trouble, especially when the data is not enough at the beginning, and the data is not enough. Our research has developed a new model, which is dedicated to recommending educational resources, the biggest feature of which is that it does not need much learning data to use.When we used a multi-modal fusion method to take into account different aspects of the information. As for the recommendation algorithm itself, we developed a prototype-based method that can be learned with a small sample to quickly adapt the model to the new situation through meta-optimization. From the experimental results, the model we proposed performed much better than the old method with very little data. Specifically, the accuracy of the recommendation was increased by 27.3%, and it was also better to improve the learning progress and coverage of knowledge.