Module 12417 (2005)

Syllabus page 2005/2006

06-12417
Nature Inspired Learning

Level 4/M

Peter Tino
10 credits in Semester 2

Links | Outline | Aims | Outcomes | Prerequisites | Teaching | Assessment | Books | Detailed Syllabus


The Module Description is a strict subset of this Syllabus Page. (The University module description has not yet been checked against the School's.)

Relevant Links

See Nature Inspired Learning Teaching Page for module material and further useful links.


Outline

This module gives an introduction to advanced learning techniques inspired by natural, physical and biological systems as well as some state-of-the-art machine learning techniques. Examples of such topics include: co-evolutionary learning, interaction between learning and evolution, the Baldwin effect, Lamarckian evolution, adaptive resonance theory neural networks, Boltzmann machines, active learning in natural computation, reinforcement learning, decision tree learning, Bayesian learning, support vector machines, mixture-model based learning.


Aims

The aims of this module are to:

  • present a range of nature inspired and state-of-the-art learning techniques
  • show how these techniques can be applied to particular problems

Learning Outcomes

On successful completion of this module, the student should be able to: Assessed by:
1demonstrate a good understanding of the nature inspired and state-of-the-art learning techniques presented in the module, as well as their relationship Examination
2demonstrate an understanding of the benefits and limitations of nature inspired learning techniques in contrast to state-of-the-art machine learning techniques Examination
3apply nature inspired and state-of-the-art learning algorithms to specific technical and scientific problems Examination

Restrictions, Prerequisites and Corequisites

Restrictions:

None

Prerequisites:

None

Co-requisites:

06-12412 (Introduction to Neural Computation) (unless 06-19341 (Introduction to Natural Computation) or equivalent has been taken previously); 06-12414 (Introduction to Evolutionary Computation) (unless 06-02411 (Evolutionary Computation) has been taken previously)


Teaching

Teaching Methods:

2 hrs/week lectures/tutorials

Contact Hours:

24


Assessment

  • Supplementary (where allowed): As the sessional assessment
  • 1.5 hr open book examination (100%).

Recommended Books

TitleAuthor(s)Publisher, Date
Machine LearningTom M. MitchellMcGraw-Hill, 1997
Neural Networks: A Comprehensive Foundation (Second Edition)Simon Haykin Prentice Hall, 1999
Evolutionary ComputationThomas Back, David B. Fogel, and Zbigniew Michalewicz (Eds.)IOP Press, 2000
Reinforcement Learning: An IntroductionRichard S. Sutton and Andrew G. BartoMIT Press, 1998
Adaptive Individuals in Evolving Populations: Models and AlgorithmsR. K. Belew and M. Mitchell (Eds.)Addison-Wesley, 1996
Fundamentals of Neural Networks: Architectures, Algorithms, and ApplicationsLaurene FausettPrentice Hall, 1994
An Introduction to Support Vector Machine and Other Kernel-Based Learning MethodsNello Cristianini and John Shawe-TaylorCambridge University Press, 2000
Pattern Classification (Second Edition)Richard O. Duda, Peter E. Hart, and David G. StorkJohn Wiley, Inc., 2001
The Elements of Statistical Learning: Data Mining, Inference, and PredictionTrevor Hastile, Robert Tibshirani, and Jerome FriedmanSpringer, 2001
The Handbook of Brain Theory and Neural Networks (2nd Edition) M.A. Arbib (Ed.)MIT Press, 2002

Detailed Syllabus

  1. Overview of what Nature Inspired Learning covers. Lessons learnt in the Neural Computation module.
  2. Neural models for processing temporal data. Recurrent neural networks and their generalizations (recursive networks). Learning in recurrent networks: Back-Propagation through time and Real-Time recurrent learning.
  3. Biologically more accurate neural models - spiking neurons. Data representation and learning in networks of spiking neurons.
  4. Introduction to probabilities and statistics. Probabilistic models of data. Classification as a probabilistic inference problem. Discriminatory vs. generative approaches to classification.
  5. Bayesian Learning -- Bayes theorem, maximum a posterior and maximum likelihood learning, Bayes optimal classifier, naive Bayesian classifier, Bayesian decision theory, Bayesian belief networks, the relationship to classical neural network models.
  6. Mixture-Model Based Learning -- mixture-model based unsupervised and supervised learning as well as EM learning algorithms, the relationship to modular neural network models. Latent-space models.
  7. Latent-space reformulations of neural topographic maps. Generative topographic mapping. Information theoretic formulations of topographic maps - channel noise models. Latent Trait models.
  8. Support Vector Machine -- generalisation in linear learning machine, large margin classifiers and learning algorithms, nonlinear support vector machines and other related issues. Kernel machines.
  9. Learning Theory - basic concepts. Complexity measures for learnng machines and their relation to bounds on generalization error.
  10. Co-evolutionary learning -- Competitive and cooperative co-evolutionary strategies and algorithms for solving learning problems and evolving artificial neural networks.
  11. Interaction between Learning and Evolution - Baldwin effect, Lamarckian evolution, how learning can guide evolution, illustrative examples on the effect of learning in evolution.

Last updated: 13 May 2005

Source file: /internal/modules/COMSCI/2005/xml/12417.xml

Links | Outline | Aims | Outcomes | Prerequisites | Teaching | Assessment | Books | Detailed Syllabus