MSc Project
In this paper we researched autoencoder based recommender systems on the SciStarter dataset. Our research covers content based recommenders, collaborative filtering recommenders using shallow and deep autoencoders, and finally hybrid recommenders.
Our results show that the SciStarter dataset is not ready to sufficiently train an autoencoder based recommender, since our experiments failed to match the benchmark. Our experiments did however show comparable results on the MovieLens dataset therefore showing that, with additional data, the SciStarter dataset could be suitable for an autoencoder based recommender systems in the future.