Wide and Deep Model

Deep Learning for Music Recommendation

Summary

In order to handle the challenge of both memorization and generalization, we introduce a new method called contented-based DNN (deep neural network) with transfer learning and pseudo-MF (matrix factorization). We apply our model to a Kaggle challenge on music recommendation and present a better result than using wide and deep model with Tensorflow APIs under these particular problem settings. We also analyze the performance of our model with a variety of other models to show the advantages and limitations of our model and propose potential work for futureresearchers

Institute
Columbia University
Date

Details

Visit my Github Repo for code and here for a final report