COMP9417 - Machine Learning

Introduction

Machine Learning comes from the field of AI. Moore's Law can be applied to data instead of transistors, and as such we need to use our increasing computational power to search through and filter this extensive data. We call this Data mining.

Machine Learning is the science of using algorithmic methods of learning from experience with the goal of improving performance on select tasks.

Data Mining is the use of ML or Statistic algorithms to search large amounts of data for hidden patterns or relationships that are interesting and potentially useful.

The two are intricately linked and near synonymous.

There are three niches with ML:

  • Data Mining
    • Using historical data to find patterns to improve/understand decisions
    • e.g. Using financial track records to compose fraud detection rules
  • Software Apps we can't program by hand
    • Such as autonomous robot agents, speech recognition or computer vision
    • e.g. Self-Driving cars
  • Self-Customising Apps
    • These are linked with the previous software apps
    • e.g. Sites that make recommendations based on user history
    • e.g. Targeted Ads

As humans we have inbuilt special knowledge, and we take and subconsciously acknowledge sensory input constantly. We must remember that computers do not have this, and have only the knowledge we grant them.

Marks

Assignments

There are two assignments in the course; a toolkit worth 10% (about week 5) and a project worth 30% (about week 12).

For the project we may propose our own idea, and may work in teams of 1-4 people.

Exam

The exam will be worth 55%.

E-Learning

There will be an online E-Learning component of the course worth 5%.

Topics

Fundamentals of Machine Learning and Data Mining

(Weeks 1-6)

Function Approximation

Computational and Statistical Foundations of Machine Learning

(Weeks 7-9)

Advanced Machine Learning Concepts and Techniques

(Weeks 10-12)