Predicting TV viewership with neural networks

Date of publication: January 15, 2017

Author: Alex Ruiz

Abstract:

Neural networks are methods widely used in tasks such as image recognition, computer vision, sequence prediction, text encoding and document classification. At Viacom's Data Science and Advanced Analytics (DSAA) we are building cutting-edge platforms that utilize a specific flavor of recurrent neural nets (Long Short-Term Memory network) that feed into broader ensemble models to predict ratings for our networks. This talk will be a blitz workshop intended for practitioners in data science with an informal account of what we've learned about the design and training of LSTMs for sequence prediction.

  • Video
  • This could also be of interest