Machine Learning on the Edge


The ProtoNN trainer can now be cloned from The trainer can be used to obtain machine learning models that are just a few kilobytes large for classical datasets such as MNIST. It also gives models in the range of a few megabytes for very large datasets with upto 100,000 labels. The repository also hosts a training algorithm for another machine learning model: Bonsai. Currently the code outputs a ProtoNN/Bonsai model in human-readable format. We plan to add support for converting the models to real prediction code for IoT devices.

EdgeML project wiki
Github repository