Numenta Newsletter — June 28, 2016
Since I arrived at Numenta 9 months ago, one of the most common requests I’ve heard is that people want an easy way to experiment with HTM without having to dive deep into our code and algorithms. Well, I’m happy to share that this week we addressed that request when we announced the release of HTM Studio, a desktop tool that allows you to test whether HTM finds interesting anomalies in your own numeric, time-series data.
The ability to detect anomalies as soon as possible in streaming data has value across a wide variety of applications – from health monitoring to preventative maintenance to geospatial tracking and more. Yet early detection is easier said than done, and traditional methods like thresholds often miss the subtle anomalies common within streaming data. Because HTM learns continuously and without supervision, it finds these types of anomalies very well.
HTM Studio provides a way for anyone, even non-technical people, to experiment with HTM for anomaly detection on their own data, without having to upload anything to the cloud. After downloading HTM Studio, you can import a local csv file, and with a few clicks, HTM analyzes your data and displays any
anomalies it finds. For those who don’t have streaming data sets, HTM Studio comes with pre-loaded sample data.
We hope that this release encourages broad experimentation with our technology. Try it on your home automation data to uncover anomalies in your usage patterns, or your sleep data to find abnormalities in your sleep-wake cycles, or your company’s sensors to see where unusual activity is occurring in the office. All you need is a csv file with one Date/Time column, one header row and at least 400 rows of numeric data listed in chronological order. For more on how to ensure your data is in the right format for HTM Studio, watch this short tutorial video. HTM Studio provides an easy way to do a proof of concept before fully implementing or deploying HTM. It also removes the technical hurdle and steep learning curve that typically accompanies HTM experimentation.
As a company with a scientific mission to understand how the brain learns, it’s
been hard for us to make this technology accessible, which is why we are so
proud of this release. If you’d like to try it, you can download the Windows or
Mac version, view our tutorials and give us feedback at
https://numenta.com/htm-studio/. Feel free to write us about
interesting anomalies that you find – we hope to share guest blog posts of the
While many of you are likely planning summer trips, Numenta will be traveling
the globe this summer as well. Our Research Engineer, Yuwei Cui, just
returned from speaking at two academic conferences in Greece (Dendrite and
Areadne), and will be in Vancouver, BC next month to speak at the
IEEE World Congress on Computational Intelligence. Subutai Ahmad, VP
of Research, is headed to speak at the Playfair AI Summit in London, before
returning to speak in San Francisco at the Data Science Summit. We’ll be
exhibiting at Data Science Summit as well, so if you’re planning to attend, be
sure to stop by our table for a demo of HTM Studio.
For those of you who follow our open source project, you may have seen that we
migrated our HTM mailing lists to the HTM Forum. With its clean,
intuitive design, this new forum provides a better platform for fostering
discussions as the community continues to grow. It categorizes content and
allows you to choose which topics to follow. Moderated by our open source flag
bearer, Matt Taylor, HTM Forum is open to anyone, so I encourage you to
read the posts and join the conversation at https://discourse.numenta.org/.
In other news, our partners have some exciting updates to share. Grok has
made several product improvements over the past few months. Grok now supports
workflow automation, features improvements to the anomaly dashboard and includes
many new design enhancements. As Grok continues to grow, the company is looking
to expand as well. If you know anyone who would make a great community manager
or full stack developer, particularly in the Southern California area, contact
Cortical.io was recently named an IDC Innovator for the
2016 Machine-Learning based Text Analytics market. Much like Numenta,
Cortical.io has been traveling the globe and will be speaking at several
upcoming events, including the AI O’Reilly Conference in New York
Lastly, we are in the final countdown of our Numenta Anomaly Benchmark (NAB)
competition. This short video explains how easy it is to win up to $5,000.