Topic: Deep Learning

1. Introduction to normalisation for computer vision
2. Introduction to the convolution operation for convolutional neural networks
3. Deep learning and abstraction
4. Introduction to convolutional and pooling layers
5. The receptive field of convolutional neural networks
6. Improving generalisation by using dropout

Theory

• Context
• Relevant NN terminology
• Complex decision boundaries: deep vs wide NNs
• The need for translation invariance
• Convolutional neural networks
• Image classification
• Quiz 10

Some things to note:

• very deep NNs can be slower.
• a very deep or wide neural network often overfit
• In practice we need to normalise the image to a range of 0 to 1 first