Using TensorFlow for letter classification by accessing properties of black-and-white, rectangular pixels. I used 2 hidden layers, applying a Sigmoid and Softmax activation fuction in both layers, respectively. For backpropogation, I used the standard Cross-Entropy loss fuction, which provides a probality for each classified letter. The Backpropogation method I used was Gradient Decent. I tested numerous values for the hyperparmeters, with the values I finally used being a learning-rate = 0.01 and Epochs = 100.
Patrick160893/Letter_Recognition----TensorFlow
Using TensorFlow for letter classification by access various black-and-white, rectangular pixels
Python