cifar-10-cnn

a convolutional neural network trained for image classification on the cifar-10 dataset

results

pre-training accuracy: 10.41%
after 10 epochs: 79.3%

confusion matrix:

python libraries used:

• pytorch
• numpy
• matplotlib
• scikit learn

ml/ai concepts:

• convolutional neural networks
• feed forward neural networks
• preprocessing
• batch normalization
• activation function (relu)
• max pooling
• dropout
• momentum-based gradient descent
• backpropogation
• cross entropy loss function