6

Abstract: I implemented a recommender system using the LightFM library. Also, the MovieLens 100K Dataset comprising 100,000 ratings from 1000 users on 1700 movies was utilized in this project. My primary purpose in this project was to compare the learning schedules of adagrad and adadelta and also four different losses. I found out the model constructed with adagrad learning_schedule and warp loss could outperform others.

See more detailes and implementation in my Github