Skip to content

Machine Learning and Predictive Analytics

Introduction

You will learn the fundamentals of Machine Learning and their application on real-world problems. The course addresses theoritical aspects of methods in Machine Learning and hands-on experience to apply your skills on problem datasets.

I recommend the following books for beginners (foundation and application):

  • Machine Learning from Scratch — Joel Grus (1st or 2nd Ed.)
  • Hands-On Machine Learning with Scikit-Learn and TensorFlow — Aurélien Géron

Resources on theory:

  • The Elements of Statistical Learning — T Hastie, R Tibshirani, J Friedman (available online)
  • Introduction to Machine Learning — Ethem Alpaydin (3d or 4th Ed.)

Structure [generic]

Week 1: Introduction
  • Introduction, aims and objectives
  • Structure
  • Toolkit, environment and language1
  • Assessment
  • Resources
Week 2: ML Concepts
  • Modelling
  • Similarity, numeric, geometric, etc.
  • Learning problems and process
  • Evaluation

Info

Sessions will involve tutorials (hands-on workshops) from Week 3 onward. Ryan (my PhD student) and Sami (AL) will be helping you.

Week 3 - 10: Methods and Learning Problems
  • Feature Engineering.
  • Supervised Learning, methods and practicals.
  • Unsupervised Learning, methods and practicals.
  • Reinforcement Learning, methods and practicals.
  • Optimisation.
Week 11: Deep Learning
  • Motives, methods and practical.
Week 12: Ethics
  • Motives and considerations

You will apply the methods you will learn, including simple KNNs, Random Forests, Logistic Rgression, Support Vector Machines, Principal Component Analysis, Deep Neural Networks and Markov Decision Process on wide range of real-world problems; in text mining, image processing, time-series analysis, regression problems, EEG, OCR, etc.

Note

Log into Blackboard for further material, including examples on previous submissions, feedback, publications of coursework, lecture notes and references/resources.

  1. Not the spoken but that of programming — I hope I'll be able to convince you with Python, less with R. ↩