COMP09012 2020 Machine Learning

Compare state of the hand engineered detectors with machine learning techniques in terms of performance on appropriate metrics and data sets and determine the appropriateness of each for safety critical applications.

Apply transfer learning to adapt a pre-trained network to a new classification problem.

Assess the validity of various cost functions to specific machine learning problems.

Effectively collaborate and communicate with others in the timely development of solutions to machine learning problems, including reports and software.

Design, test and evaluate deep network architectures.

Appreciate the data rights of citizens and the constraints these apply to the use of pattern detection in real world scenarios.

A lecture will be provide each week. In advance of the lecture, the learner will be asked to review key text book chapters and academic papers that are relevant to the lecture so that they get the maximum learning from that lecture.

The Project work, both team and individual, will challege the learner to master concepts beyond those covered in the theory lecture. This will prepare them for the life long learning that will be required in the fast moving field of Computer Vision.

A terminal exam and continuous assessment in the form of group project work will be used to assess the module.

To reinforce the theoretical principles covered in lectures, learners will participate in project work.

The learner will complete a final exam at the end of the semester.

The learner is required to pass both the projects and terminal examination element of this module.

Repeat Exams will be set for Autumn of each year.

Repeat project work can be submitted at the repeat exam sitting.

Machine Learning: Logistic Regression, KNN, SVM, kernel SVM, Decision Trees.

Principle Component Analysis (PCA)

Model evaluation and Hyperparameter Tuning.

Deep Learning: Perceptron, multi-layer perceptron, representation learning.

Back propagation, SGD (variants), Cost/loss functions, regularisation.

Convolutional Neural Networks.

Dropout, Batch-normalisation.

Image classification, image recognition, localisation.

Transfer Learning.

Introduction to Reinforcement Learning.

Data Sets, ImageNet, Kaggle.

Data distributions, bias, variance, divergence (KL, JS, Wasserstein Distance).

Machine Learning Frameworks such as TensorFlow, Keras, PyTorch and Caffe.

Citizen rights under data protection, data storage.