Netlab Book: Worked Examples

Each chapter in the book contains one or more worked examples illustrating how Netlab can be used and extended. The software that implements the worked examples given in the book are available as zipped files from the following links (or the whole lot in one zip file). Note that the solutions to the exercises are not available!

• Chapter 1: Data normalisation

• Chapter 2: Constrained optimisation

• Chapter 3: Applications of mixture models

This file contains the two scripts (demgmmc.m and novel.m for classification and novelty detection respectively). It also contains an elaboration of the classification algorithm turning it into a proper Netlab model: constructor gmmc, training algorithm gmmcem and forward propagation gmmcfwd. The second demonstration program demgmmc2 shows how this works in practice.

• Chapter 4: Training regularised models

• Chapter 5: Direct connections

• Chapter 6: Dual basis functions and network complexity

• Chapter 7: Canonical variates

Function canvarand and test script testcanvar.

• Chapter 8: Convergence diagnostics

Function convcalc and demonstration demconv2.

• Chapter 9: Improved classification approximation

• Chapter 10: Gaussian Processes for classification

Functions described in text together with demonstration demgpc.

Download the software and unzip it in a separate directory from the main Netlab installation to avoid over-writing Netlab files. The files are in Unix format, but most PC based editors can cope (in particular, the Matlab editor can read the files correctly).