The Case for Sparsity in Neural Networks | Numenta News – October 2020

Christy Maver • VP of Marketing

In this month’s newsletter, we have an update on our machine learning work and information on upcoming events.

Numenta Machine Learning Updates: Performance Improvements and Dynamic Sparse Networks

Over the past several months, we’ve made great progress in our experimentation of applying neuroscience principles to today’s machine learning and AI systems. Specifically, we have been focused on applying sparsity to neural networks. Sparsity is foundational to the brain’s ability to run efficiently. In machine learning, sparsity offers the promise of significant performance benefits and energy savings. We’ll have more to share about Numenta’s results on performance improvements in sparse networks over the coming weeks.

In addition to performance improvements, one of the areas we are exploring is how to create highly sparse networks that learn their structure dynamically through training. This approach is known as dynamic sparsity. In a new post titled, “The Case for Sparsity in Neural Networks Part 2: Dynamic Sparsity,” researcher Lucas Souza continues his literature review series on sparse neural networks. He provides a review of current dynamic sparse algorithms and discusses the implications to hardware. Check out the blog post and stay tuned for the third and final post in this series, which will focus on sparse activations.

Numenta Events

From Neuroscience to Artificially Intelligent Systems (NAISys) – Nov 9-12
Our Co-Founder Jeff Hawkins and VP of Research and Engineering Subutai Ahmad are scheduled to present at the NAISys conference organized by Cold Spring Harbor Labs November 9-12. The goal of this virtual meeting is to bring together researchers at the intersection of AI and neuroscience, and to identify insights from neuroscience that can help catalyze the development of next-generation artificial systems.

Jeff will deliver a presentation on Numenta’s Thousand Brains Theory of Intelligence – our theory of how the brain works and how it creates intelligence. Pulling from his upcoming book, A Thousand Brains, Jeff will discuss the key components of the theory, as well as how the theory will shape the future of artificial intelligence. Subutai will present a poster, “Sparsity in the neocortex, and its implications for machine learning.” Registration for the conference is still open, and we plan to share the materials on our website after the conference.

Brain Inspired Computing Congress – November 18-19

Subutai will speak at the Brain Inspired Computing Congress November 18-19. This on-demand only, virtual event was designed to bridge the gap between research and commercial application of brain-inspired technologies. Topics will range from new architectures for neuromorphic chips to current efforts underway for creating biologically plausible algorithms. Subutai’s presentation will cover our work on applying sparsity to neural networks and results on various performance metrics.

Stay tuned for further updates on events, including our next Brains@Bay Meetup, and visit our YouTube channel for videos of past meetups, events, and research meetings.

Thank you for continuing to follow Numenta.

Christy Maver
VP of Marketing

Christy Maver • VP of Marketing

Leave a Reply

Your email address will not be published. Required fields are marked *