Mnist Dataset Comparison Example?

General Loihi question: I was recently told there was a performance-benchmarking simulation example somewhere on the Nengo website that demonstrated how to compare CPU, GPU, and Loihi performance by using both numerical and fashion MNIST datasets. I have searched the website and cannot find this simulation example. So my questions are: does such a simulation example exist, if so, where would this example be located, and could you please point me in that direction?

Thanks in advance.


I think you might be referring to our arXiv paper where we run a keyword spotting network on various devices and compare the performance and power usage across various device. The repository for this paper can be found here, and there are different branches in the repository for the different versions of the models run on the different devices.

I took a quick look through the repository, and all of it seems to be there except for the Loihi code (I can’t remember whether that code has been uploaded to that specific repo), but a simpler version of the Loihi network can be found on our NengoLoihi documentation (or in the NengoLoihi pip package).

We have multiple Jupyter notebook examples that use the numerical and fashion MNIST data sets. I’ve listed some of them below:

Thank you Xchoo! I’ve seen and run the last three examples you listed. I really appreciate your continued support.

Have a great weekend!