Accessibility Skip to Global Navigation Skip to Local Navigation Skip to Content Skip to Search Skip to Site Map Menu

COSC420 Neural Networks

Introduction to neural networks - computational tools inspired by the brain - which give a different perspective on the nature of computation and complex topics such as vision, language, learning and memory.

Despite its slow "hardware," the brain is a much more powerful and sophisticated computational system than any computer ever built. What can the brain teach us about computation and how to perform complex tasks, such as natural language processing, vision and control and optimisation problems? Artificial neural networks are a family of methods that try to address these issues and explore "brain-like" computation, information processing and learning. "Deep learning" is a currently popular technology based on neural networks and used by organisations such as Google, Baidu, Microsoft and Apple.

Paper title Neural Networks
Paper code COSC420
Subject Computer Science
EFTS 0.1667
Points 20 points
Teaching period First Semester
Domestic Tuition Fees (NZD) $1,307.76
International Tuition Fees (NZD) $5,517.77

^ Top of page

Restriction
NEUR 420
Eligibility

There are no formal prerequisites for the 400-level papers, but prior knowledge is assumed.

Contact

Computer Science Adviser (adviser@cs.otago.ac.nz)

Teaching staff
Lecturer: Professor Anthony Robins
Paper Structure
As the topic of artificial neural networks will be new to most, the paper is divided into two phases. The first introduces basic material and will be in a lecture-driven format. We will cover an introduction to the topic, the practical use and tuning of neural nets and fundamental algorithms and architectures, including (a selection of):
  • 1-layer nets
  • Multi-layer nets
  • Back propagation
  • Hopfield nets
  • Boltzmann machines
  • Competitive learning
  • Self-organising maps
  • Recurrent nets
  • Dynamic architectures
  • Reinforcement learning
The second phase covers current research within the field and will be in a more open, student-driven format. The material will be different every year, as it will be driven by students' specific interests. Topics which have been frequently covered in the past include:
  • Advanced neural network theory
  • Applications to vision and robotics
  • Purpose-built hardware ("neurocomputers")
  • Applications within artificial life and software agents
  • Neural/symbolic hybrids
  • Mathematical and Baysean interpretations
  • Applications to neuroscience and psychological modelling
Assessment:
  • Assignment 25%
  • Presentation and reviews 10%
  • Quizzes and participation 5%
  • Final exam 60%
Teaching Arrangements
One session per week. Sessions include a mixture of quizzes, lecture-type presentation of new material, tutorial-type discussion sessions and student presentations.
Textbooks
Text books are not required for this paper.
Course outline
View the course outline for COSC 420
Graduate Attributes Emphasised
Interdisciplinary perspective, Lifelong learning, Communication, Critical thinking, Research.
View more information about Otago's graduate attributes.
Learning Outcomes
This paper will enable students to:
  • Understand and implement a range of artificial neural networks
  • Understand the strengths and weaknesses of neural networks compared to traditional symbolic methods
  • Develop an interdisciplinary perspective incorporating computer science, psychology and neuroscience
  • Perform practical research with a neural network simulation and systematically present the outcomes
  • Perform a literature review and evaluate publications in the field
  • Present a summary of recent literature in a seminar format

^ Top of page

Timetable

First Semester

Location
Dunedin
Teaching method
This paper is taught On Campus
Learning management system
None

Lecture

Stream Days Times Weeks
Attend
A1 Tuesday 09:00-10:50 9-15, 18-22