Sitting at the intersection of neuroscience and computer science, we encounter interest in our work from a variety of people. For the scientific and academic community, we have a growing list of peer-reviewed papers. The machine learning and data science enthusiasts can dive into our open source code and experiment with it firsthand. But what about those of us with no exposure to neuroscience – those of us who are neither mathematicians nor engineers, those of us who’ve never taken a computer science course? Where do we go when we want to understand the fundamental concepts of HTM Theory? We go to HTM School.
HTM School was created nearly a year ago, by Numenta’s Open Source Flag Bearer, Matt Taylor, in an
attempt to make HTM Theory more accessible by breaking it down into bite-sized, 15-20 minute episodes filled with stories, examples and illustrations. The series is designed to be viewed in order, starting with a high level overview in Episode 0. From there, Matt lays out core components of the theory, one episode at a time: from how we represent information to how we learn spatial patterns.
The series has attracted a nice following, with more than 35,000 total views and a healthy amount of Q&A that takes place in the comments section and on the HTM Forum.
Recently, Matt published the 12th episode in the series, “Temporal Memory, Part 1,” marking an educational milestone as the first episode to introduce the concept of sequence memory. Readers of On
Intelligence will remember this concept of how the brain remembers sequences of events and makes predictions based on those memories. In “Temporal Memory, Part 1,” Matt describes in detail how neurons remember sequences within the context of previous inputs, and make predictions based on this stimulus.
These are not easy concepts, to be sure. Understanding the neuroscience underlying HTM theory can be difficult, especially for novices. Yet the implications of understanding how the brain learns and implementing these concepts in software are significant. With HTM School, Matt peels back the layers of complexity and uses computer animated 3D visualizations and examples to make HTM theory accessible
If you haven’t been to HTM School yet, I recommend starting at Episode 0, and working your way up to the most recent episode. Matt’s intention is to continue to produce more videos, so now’s the perfect time to get caught up. If you have a question about the content, sound off in the comments of the videos or post your questions on the HTM Forum, where Matt has a dedicated section for discussion associated with these videos.
Jeff and Subutai to deliver opening keynote at Cornell Silicon Valley on March 7 in San Francisco
In between your HTM School tutorials, for those of you in the San Francisco area, you can catch Jeff Hawkins and Subutai Ahmad deliver the opening keynote at Cornell Silicon Valley on March 7 at 1pm: “Reverse-engineering the brain for intelligent machines.” Jeff and Subutai will have a conversation about the work we do at Numenta and the progress we’ve made. Don’t miss this rare opportunity to see them