Representation Power of Neural Networks

Channel:
Microsoft Research
Subscribers:
344,000
Published on ● Video Link: https://www.youtube.com/watch?v=byHU2Vlp2Vs


Duration: 59:01
1,057 views
9

This talk will survey a variety of classical results on the representation power of neural networks, and then close with a new result separating shallow and deep networks: namely, there exist classification problems where any shallow network needs exponentially as many nodes to match the accuracy of certain deep or recurrent networks.



Other Videos By Microsoft Research


2016-06-13Machine Learning as Creative Tool for Designing Real-Time Expressive Interactions
2016-06-13Recent Developments in Combinatorial Optimization
2016-06-13Computational Limits in Statistical Inference: Hidden Cliques and Sum of Squares
2016-06-13Coloring the Universe: An Insider's Look at Making Spectacular Images of Space
2016-06-13Towards Understandable Neural Networks for High Level AI Tasks - Part 6
2016-06-13The 37th UW/MS Symposium in Computational Linguistics
2016-06-13The Linear Algebraic Structure of Word Meanings
2016-06-13Machine Learning Algorithms Workshop
2016-06-13Interactive and Interpretable Machine Learning Models for Human Machine Collaboration
2016-06-13Improving Access to Clinical Data Locked in Narrative Reports: An Informatics Approach
2016-06-13Representation Power of Neural Networks
2016-06-13Green Security Games
2016-06-13e-NABLE: A Global Network of Digital Humanitarians on an Infrastructure of Electronic Communications
2016-06-10Microsoft Research New England: An introduction
2016-06-06Python+Machine Learning tutorial - Data munging for predictive modeling with pandas and scikit-learn
2016-06-06Symposium: Deep Learning - Xiaogang Wang
2016-06-06Symposium: Deep Learning - Leon Gatys
2016-06-06Symposium: Brains, Minds and Machines - Surya Ganguli
2016-06-06Posner Lecture: Probabilistic Machine Learning - Foundations and Frontiers
2016-06-06Invited Talk: Learning with Intelligent Teacher: Similarity Control and Knowledge Transfer
2016-06-06Tutorial: High-Performance Hardware for Machine Learning


Tags:
microsoft research
neural networks
computer systems and networking