This project aims to classify the Cifar10 dataset using a three-layered Convolutional Neural Network (CNN). Cifar10 dataset consists of 60000 images which represent 10 classes. For more information, check this website. https://www.cs.toronto.edu/~kriz/cifar.html In order to find the best CNN, we change some of the following factors:

1. Number of hidden layers (from 0 to 2).
2. Different types of activation functions.
3. Different optimizers.

In the next step of this project, we investigate the effect of the size of the input dataset. (we reduce the dataset so that it would have 600 images per class)

Finally, we add dropout to the network and see its effect on the classification.