ML

This contains the basic programmes as a part of my ML learning phase, including the forward propagation , back propogation, basic networks of CNNs and ANNs and two other real life prediction examples. One of the programs contains Predictions of handwritten digits, the dataset loaded from MNIST. Initially when one hidden layer was taken with 128 neurons and 5 epochs giving accuracy on test set as 97.72% Next on increasing epochs to 20 accuracy increased slightly to 97.97% as there is always a certain value of epoch when we get best accuracy,beyond which overfitting may take place. In next case, on adding one more hidden layer with 100 neurons gave the best accuracy as 97.98% as addition of hidden layers helps in better learning upto a certain limit.