{"id":442986,"date":"2017-11-28T05:17:49","date_gmt":"2017-11-28T13:17:49","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-project&p=442986"},"modified":"2017-11-29T09:01:38","modified_gmt":"2017-11-29T17:01:38","slug":"stochastic-neural-networks","status":"publish","type":"msr-project","link":"https:\/\/www.microsoft.com\/en-us\/research\/project\/stochastic-neural-networks\/","title":{"rendered":"Stochastic Neural Networks"},"content":{"rendered":"
When we write a letter, have a conversation, or draw a picture, we exercise a uniquely human skill by creating complex artifacts that embody information. Current AI technology cannot yet match human ability in this area because it fails to have the same understanding of the world in terms of independent causal factors. Without such understanding AI systems cannot compose information in the rich ways a human could.<\/p>\n
Instead, current successful AI technology requires a large amount of supervised data in order to learn different plausible combinations of independent variations.\u00a0In the stochastic neural network project we aim to build the next generation of deep learning models which are more data-efficient and can enable machines to learn more efficiently and eventually to be truly creative.<\/p>\n
Research in the stochastic neural networks project addresses this research challenge along three lines:<\/p>\n
We analyze and improve important algorithms for generative models such as methods for generative adversarial networks, models based on integral probability metrics, and variational autoencoders.<\/p>\n