IDAE: Imputation-boosted Denoising Autoencoder for Collaborative Filtering

摘要: In recent years, while deep neural networks have shown impressive performance to solve various recognition and classification problems, collaborative filtering (CF) received relatively little attention utilize networks. Because of inherent data sparsity, it remains a challenging problem for this paper, we propose new CF model, namely the imputation-boosted denoising autoencoder (IDAE), top-N recommendation. Specifically, IDAE consists two steps: imputing positive values learning with imputed values. First, infers imputes user feedback from missing Then, correlation between items is learned by using (DAE) Unlike existing DAE that randomly corrupts input, key characteristic original are taken as reflected corrupted output. Our experimental results demonstrate significantly outperforms state-of-the-art algorithms autoencoders (by up 5%) on MovieLens datasets.