jq-neural-network A neural network written completely in jq. Sample program processes the MNIST image dataset with a success rate of 94%. Running the sample program: This will first compile make_json.c, which parses MNIST data files and produces JSON. Then it will run make_json and pipe it into jq, first processing the training set and then the test set. When finished, jq will emit the error rate of the test set. This usually takes days to run to completion. How it works A typical feedforward n...