IAPR Techical Committee 3: Neural Networks and Machine Learning

[IAPR logo]

The scope.

TC3 is one of the Technical Committees of the International Association for Pattern Recognition, IAPR. The scope of the Technical Committee 3 is not only neural networks, but also what is presently called Computational Intelligence, including at least artificial neural networks, fuzzy systems, evolutionary computing like genetic algorithms, and those branches of statistics that are relevant to the above fields. However, the emphasis is on classification problems rather than general function approximation etc., to make our profile different from and sharper than computational intelligence in general. Some excellent textbooks have recently appeared, highlighting the connections of neural nets with mainstream statistical classification methods.

Main activities.

Comparisons and benchmarking of classification methods, concentrating on machine learning and neural networks, is the main theme of the Committee. Comparison studies have been reported, especially on OCR, but they do not yet address the problems of a non-expert user working on an application and wanting some good results. He/she wants to know what classifier to use and how exactly to choose the various parameters, especially in neural network algorithms like back-prop. It has turned out that similar efforts are underway elsewhere, and contacts have been made. A reference list of available benchmarking databases and some test results are included below on this page. This list will be complemented by pointers to new ones when they become available.

Workshop activity is one of the main tasks of any TC. However, the field of artificial neural networks has grown very large and diverse, and there are so many neural network conferences annually all over the world, that it has been felt in the Committee that there is presently no significant market for a conference of our own. Good contacts to the program and organizing committees of the existing major conferences (like those organized by IEEE, INNS, and ENNS) are maintained by individual members of the TC3, and in this way the PR aspect can be strengthened in those meetings.


The TC now has 26 members listed. The members are presently:

How to join the TC3?

Please send your name and address, including the email, to Ari Visa, ari.visa@cs.tut.fi.

Databases and Benchmarks

Below is a list of some public domain databases, classification benchmarks, and related information in the web:




Project StatLog (Esprit Project 5170) archives


ELENA (Esprit Project 6891) archives


Delve datasets, utilities, benchmarks


UCLA List of available data sets


comp.ai.neural-networks FAQ list of databases

Last modified: Fri Jun 16 10:59:18 EET 2000Mail to maintainer: ari.visa@cs.tut.fi