Timezone: »

Learning to Combine Top-Down and Bottom-Up Signals in Recurrent Neural Networks with Attention over Modules
Sarthak Mittal · Alex Lamb · Anirudh Goyal · Vikram Voleti · Murray Shanahan · Guillaume Lajoie · Michael Mozer · Yoshua Bengio

Thu Jul 16 12:00 PM -- 12:45 PM & Fri Jul 17 12:00 AM -- 12:45 AM (PDT) @ None #None

Robust perception relies on both bottom-up and top-down signals. Bottom-up signals consist of what's directly observed through sensation. Top-down signals consist of beliefs and expectations based on past experience and the current reportable short-term memory, such as how the phrase `peanut butter and ...' will be completed. The optimal combination of bottom-up and top-down information remains an open question, but the manner of combination must be dynamic and both context and task dependent. To effectively utilize the wealth of potential top-down information available, and to prevent the cacophony of intermixed signals in a bidirectional architecture, mechanisms are needed to restrict information flow. We explore deep recurrent neural net architectures in which bottom-up and top-down signals are dynamically combined using attention. Modularity of the architecture further restricts the sharing and communication of information. Together, attention and modularity direct information flow, which leads to reliable performance improvements in perceptual and language tasks, and in particular improves robustness to distractions and noisy data. We demonstrate on a variety of benchmarks in language modeling, sequential image classification, video prediction and reinforcement learning that the \emph{bidirectional} information flow can improve results over strong baselines.

Author Information

Sarthak Mittal (Uber ATG)
Alex Lamb (Universite de Montreal)
Anirudh Goyal (Université de Montréal)
Vikram Voleti (Mila, University of Montreal)

## Vikram Voleti I am a PhD student at Mila, University of Montreal, I work in computer vision and deep learning. My supervisor is Prof. Chris Pal. More details on my website - https://voletiv.github.io

Murray Shanahan (DeepMind / Imperial College London)
Guillaume Lajoie (Mila, Université de Montréal)
Michael Mozer (Google Research / University of Colorado)
Yoshua Bengio (Montreal Institute for Learning Algorithms)

More from the Same Authors