A neural network to classify the MNIST collection of handwritten numbers using TensorFlow. The version 1 network, created on June 19, 2020 runs at about a 90% success rate, which is less than the average rate for human comprehension. See Wikipedia for details:
https://en.wikipedia.org/wiki/MNIST_database#Classifiers
Version 2 was created on June 22, 2020, and has a much better accuracy of 98%. Some improvements I made were to include dropouts, regularize the data, and use activation functions.