NeMo is a simulation model designed to allow experimentation with different neuromorphic hardware. Built on top of ROSS, NeMo features excellent scalability across HPC clusters. NeMo is targeted at current neuromorphic hardware designs, and as such currently simulates spiking neural networks in a grid layout, similar to the IBM TrueNorth Processor.
Currently, NeMo is in a state of development. The latest working version of NeMo (0.1.0) is available in the branch. The latest master version is a new implementation that will feature improved memory performance along with high performance I/O. The next release will feature more common spiking neural network
Benchmarks and results of the first edition of NeMo is available at the ACM Digital Library.
Version 0.1.0 supports generating probabilistic networks for benchmarking purposes. It also only supports the IBM TrueNorth neuron specification, available here.
