THE UNREASONABLE EFFECTIVENESS OF A TWO-LAYER NEURAL NETWORK FOR RECOMMENDATIONS

Anonymous

Nov 03, 2017 (modified: Nov 03, 2017) ICLR 2018 Conference Blind Submission readers: everyone Show Bibtex
  • Abstract: We present a personalized recommender system using neural network for recommending products, such as eBooks, audio-books (Audible), Mobile Apps, Video and Music. It produces recommendations based on user consumption history: purchases, listens or watches. Our key contribution is to formulate recommendation problem as a model that encodes historical behavior to predict the future behavior using soft data split, combining predictor and auto-encoder models. We introduce convolutional layer for learning the importance (time decay) of the purchases depending on their purchase date and demonstrate that the shape of the time decay function can be well approximated by a parametrical function. We present offline experimental results showing that neural network with two hidden layers can capture seasonality changes, and at the same time outperforms other modeling techniques, including our recommender in production. Most importantly, we demonstrate our model with two hidden layers can be scaled to all digital categories. Finally, we show online A/B test results, discuss key improvements to the neural network model, and describe our production pipeline.
  • TL;DR: Improving recommendations using time sensitive modeling with neural networks in multiple product categories on a retail website
  • Keywords: Recommender systems, deep learning, personalization

Loading