I’d like to take the time to introduce to you a new feature which I’ve added to the community fork of Nupic. Although the majority of of the library is still under development, this piece is ready for the spotlight.
Sparse Distributed Representations (SDRs) are a critical component of all HTM models, and are used throughout Nupic. Numenta has done significant mathematical analysis of SDRs and their properties.
I made a new class SDR for working with Sparse Distributed Representations. The SDR class has methods to hold, measure, and manipulate them. Key features include:
- Seamless conversions between all commonly used SDR data formats (sparse, dense, coordinates)
- Methods to generate random SDRs & add noise to SDRs
- Methods to measure Sparsity, Overlap, Activation Frequency & Binary Entropy
- Integration with the other HTM algorithms
- Written in C++, usable in python via bindings.
For more information see: https://github.com/htm-community/nupic.cpp/wiki/Sparse-Distributed-Representations
Questions and comments are welcome.