Numenta

Numenta's machine intelligence technology is called hierarchical temporal memory (HTM), and is a computational theory of the neocortex.[4] This theory was first described in the book On Intelligence, written in 2004 by Jeff Hawkins and co-author Sandra Blakeslee. At the core of HTM are time-based learning algorithms that store and recall temporal patterns. The HTM algorithms are documented and available through its open source project, NuPIC. The HTM technology is suited to address a number of problems, particularly those with the following characteristics: streaming data, underlying patterns in data change over time, subtle patterns, time-based patterns.[5]

Numenta focuses on large-scale brain theory and simulation. Numenta researchers work with experimentalists and publish results to derive an understanding of the neocortex. Their main research focus areas are cortical columns, sequence learning and sparse distributed representations. In October 2018, they released a major theory for intelligence and cortical computation called The Thousand Brains Theory of Intelligence.[6] They have written a number of peer-reviewed journal papers and research reports on these topics.

Numenta is a technology provider and does not create go-to-market solutions for specific use cases. The company licenses their technology and application code to developers, organizations and companies who wish to build upon their technology."[7] Numenta has several different types of licenses, including open source licenses, trial licenses and commercial licenses. Developers can use Numenta technology within NuPIC using the AGPL v3 open source license.

The following commercial applications are available using NuPIC:

• Grok – anomaly detection for IT servers.
• Cortical.io – advanced natural language processing.

The following tools are available on NuPIC:

• HTM Studio – find anomalies in time series.
• Numenta Anomaly Benchmark – compare HTM anomalies with other anomaly detection techniques.

The following example applications are available on NuPIC:

• HTM for Stocks – example of tracking anomalies in the stock market (sample code)
• Rogue behavior detection – example of finding anomalies in human behavior (white paper and sample code)
• Geospatial tracking – example of finding anomalies in objects moving through space and time (white paper and sample code)

Numenta works with strategic partners, who license their technology and build products using HTM. Cortical.io is using HTM for natural language processing, the partnership was announced in May 2015.[8] Avik Partners has licensed their Grok for IT Analytics application to monitor IT servers.[9] The partnership was announced in August 2015. Numenta also partners with various research institutions and universities.

The Numenta Platform for Intelligent Computing is an open source platform and community for machine intelligence based on HTM theory. NuPIC is an implementation of HTM and can be used to analyze streaming data. Numenta first announced in June 2013 that it would open-source its HTM technology, the core of its software and algorithms. This was accompanied by the new Numenta.org website[10] and a mailing list for community members.

Community members are contributors from around the world, and topics on the mailing list have included both discussions of the HTM theory and details of software development. The mission of NuPIC is to build and support a community that is interested in machine learning and machine intelligence based on modeling the neocortex and its principles.[11]

Numenta has hosted a series of hackathons,[12] the first one in 2013, to bring community members together to collaborate on NuPIC and its applications.

NuPIC codebase is based on 2.7.x Python that is scheduled to be obsoleted in 2020.[13][14] There is no active plan to update it to support a modern version of Python at the moment.[13] As of 2019, a PyTorch version of NuPIC remains in unsteady development.[15]

• Official website