Mnist_DeepLearning

• This repository is dedicated to the 'hello world' of Deep learning, the MNIST dataset.

Objective

• Explore the MNIST dataset and try to achieve 98% + accuracy on the test set.

Final Result

• Model used the hyperparameters :

  • Hidden_layer_size = 200
  • Depth : 3 hidden layers
  • Relu activation function
  • Optimizer : Adam optimizer
  • Batch_size : 100

• Validation accuracy : 99.38%

• Test accuracy : 99.48%