KF5012: Introduction to Artificial Intelligence

Lecture 5: Convolutions and Engineering

Part 1: Recap and Code Structure

Part 2: Convolutional Neural Networks

This video introduces the idea of a convolution and explains why convolutional layers can be preferable to fully connected layers on certain types of data

Part 3: Activation and Loss Functions

There are two missing piece of neural networks that we haven’t yet talked about: activation functions and loss functions. This video explores both in detail.

Part 4: Hyperparameters and Dropout

This video introduces some important hyperparameters before going into detail about Dropout, one of the most important hyperparameters to protect against overfitting

Part 5: Batch Normalisation and Convolution Details

This video explores the need to normalise at different layers of your network, before explaining some of the finer points of convolutional layers

Part 6: Data Augmentation

Deep Neural Networks are hungry for data. This video describes how to make the most out of the available data to improve your performance

Part 7: Putting it all Together

This video takes the concepts introduced over the last two lectures and demonstrates how to combine them into an easy to code Keras model