NuPIC is a library that provides the building blocks for online prediction and anomaly detection systems. The library contains the Cortical Learning Algorithm (CLA), but also the Online Prediction Framework (OPF) that allows clients to build prediction systems out of encoders, models, and metrics.

For more information, see numenta.org or the NuPIC wiki.

For more detailed documentation, see the OPF wiki page.

Encoders turn raw values into sparse distributed representations (SDRs). A good encoder will capture the semantics of the data type in the SDR using overlapping bits for semantically similar values.

Models take sequences of SDRs and make predictions. The CLA is implemented as an OPF model.

Metrics take input values and predictions and output scalar representations of the quality of the predictions. Different metrics are suitable for different problems.

Clients take input data and feed it through encoders, models, and metrics and store or report the resulting predictions or metric results.

For all installation options, see the Installing and Building NuPIC wiki page.

Currently supported platforms:

• Linux (32/64bit)
• Mac OSX
• Raspberry Pi (ARMv6)
• Chromebook (Ubuntu ARM, Crouton) (ARMv7)
• VM images

Dependencies:

• Python (2.6-2.7) (with development headers)
• GCC (4.6-4.8), or Clang
• Make or any IDE supported by CMake (Visual Studio, Eclipse, XCode, KDevelop, etc)

The dependencies are included in platform-specific repositories for convenience:

• nupic-linux64 for 64-bit Linux systems
• nupic-darwin64 for 64-bit OS X systems

Complete set of python requirements are documented in requirements.txt, compatible with pip:

pip install -r external/common/requirements.txt

Note: If using pip 1.5 or later:

pip install --allow-all-external --allow-unverified PIL --allow-unverified psutil -r external/common/requirements.txt

Note: If you get a "permission denied" error when using pip, you may add the --user flag to install to a location in your home directory, which should resolve any permissions issues. Doing this, you may need to add this location to your PATH and PYTHONPATH. Alternatively, you can run pip with 'sudo'.

Set the following environment variables in your ~/.bashrc file. $NUPIC is the path to your NuPIC repository and $NTA is the installation path for NuPIC. You may set a different path for $NTA or specify the location with CMake with the command line option -DPROJECT_BUILD_RELEASE_DIR:STRING=/my/custom/path.

export NUPIC=<path to NuPIC repository>
export NTA=$NUPIC/build/release
export PYTHONPATH=$PYTHONPATH:$NTA/lib/python<version>/site-packages

mkdir -p $NUPIC/build/scripts
cd $NUPIC/build/scripts
cmake $NUPIC

cd $NUPIC/build/scripts
make -j3

Note: -j3 option specify '3' as the maximum number of parallel jobs/threads that Make will use during the build in order to gain speed. However, you can increase this number depending your CPU.

cd $NUPIC/build/scripts
# all C++ tests
make tests_everything
# C++ HTM Network API tests
make tests_cpphtm
# Python HTM Network API tests
make tests_pyhtm
# Python OPF unit tests
make tests_run
# Python OPF unit and integration tests (requires mysql)
make tests_run_all
# Run all tests!
make tests_all

• Open CMake executable.
• Specify the source folder ($NUPIC).
• Specify the build system folder ($NUPIC/build/scripts), i.e. where IDE solution will be created.
• Click Generate.
• Choose the IDE that interest you (remember that IDE choice is limited to your OS, i.e. Visual Studio is available only on CMake for Windows).

• Open nupic.*proj solution file generated on $NUPIC/build/scripts.
• Run ALL_BUILD project from your IDE.

• Run any tests_* project from your IDE (check output panel to see the results).

For examples, tutorials, and screencasts about using NuPIC, see the Using NuPIC wiki page.