COMP4702 Lecture 1

Lecture notes regarding the organisation of the course.

slides

Marcus has introduced some background material, from which I have taken some notes:

Computational Complexity | Statistics and Probability | Calculus and Linear Algebra

Table of Contents

Introduction to ML

Course Admin

Lindholm - Chapter 1

Intro to ML

What is Machine Learning

• A subfield of Artificial Intelligence
  • Goal: To create computer systems that are intelligent (strong), or fool you into thinking they are (weak)
  • One thing we associate with (natural) intelligent entities: adaptive behaviour, and/or the ability to learn.
  • So machine learning is about creating computer systems that can learn to perform tasks.
• In other words, machine learning software automatically builds software to perform a task
  • On the basis of data.
  • So machine learning is about engineering algorithms that learn from data to solve interesting problems.
  • The study of computer algorithms capable of learning to improve their performance on a task on the basis of their own experience.
  • This is sometimes called inductive learning
• Learning is an extremely broad concept, studied by psychologists, neuroscientists and others.
• Machine Learning restricts itself to a few well-defined classes of learning problems that are still very general.
  • Data-centred learning.
• Furthermore, we concentrate mainly on problems that involve numerical data, or at least “structural” data.
  • Nothing to do that involves language comprehension or high level cognitive-type stuff
  • Our problems might be called sub-symbolic or low-level.
  • structural data meaning that the input is converted to numerical data in some form.

The Rise of the (Learning) Machines

• Machine Learning (incorrectly aka Artificial Intelligence) is one of the hottest topics on the planet.
• The surge in interest and attention in the last few years has been fast, frantic and exciting.
• What is going on?
  • The world has built a densely connected network of computers and devices which is widely available and relatively cheap.
  • Technology has lead to an explosion in the collection, storage and availability of data (hence, data science, data analytics).
  • There have been some big improvements in ML applications; tech giants have thrown huge amount of $ at AI/ML, everyone has started getting excited (and maybe afraid…)

The Big Questions

• How far can AI go? (Persona assistants, healthcare, self-driving cars, autonomous robots, language understanding)
• What happens if technology displaces a large chunk of the labour force?
• Autonomous weapons?
• Data security?
• Conscious machines?
• The singularity?

The Price of Fame and Fortune?

• Currently huge money, interest and effort in AI
• Scientific rigour is incredibly important.
• The objectives and applications of AI connect directly to issues of humanity such as ethics, bias, explainability, safety, trust, etc.
• With critical mass, these factors can lead to pretty volatile situations and drama that computer scientists would not normally be involved in!

Course Admin

• Using Lindhurst textbook here.

Assessment

Demo: 2 demos in the semester, a timetable for when your demo is will be released at some point

• Present your demo classwork live to a tutor in the prac (10 to 15 minutes)
• Start in week 3/4
• You can demo anything from a previous prac, don’t need to complete every prac
• Considering marking pracs as pass/fail.

Report: Due on the last day of the semester; Have the entire semester to work on the assignment.

• Checkpoint for the assignment half way through (around wk7/8) and get feedback on your assignment.

Final Exam: Going back to paper-based final exam instead of take-home assignment format.

• Similar to 2019-era style exam - short answer, calculations, longer response questions