Argument-based Machine Learning

Description

Standard machine learning takes examples as input in the form of pairs (A, C), where A is an attribute value vector and C the class the example belongs to e.g., (Name=Mrs.Brown, PaysRegularly=no, Rich=yes, HairColor=Blond)^T, C=(CreditApproved=Yes).

In contrast, argument-based machine learning (ABML) is able to process examples of the form (A, C, Arguments), where Arguments is a set of type

• C because of reasons, or
• C despite reasons.

Example

• A = (Name=Mrs.Brown, PaysRegularly=no, Rich=yes, HairColor=Blond)^T
• C = (CreditApproved=Yes)
• Arguments = (C because of Rich, C despite PaysRegularly = No)

Argument-based machine learning has so far been applied to

• CN2 (Možina, M. et al., 2008) - ABCN2 and
• Support Vector Machines (SVM) (Lucas Carstens, 2011)

Related Resources

1. Lucas Carstens, 2011, Sentiment Analysis - A multimodal approach, Master Thesis, Imperial College, London
2. Možina, M. et al., 2008. Fighting Knowledge Acquisition Bottleneck with Argument Based Machine Learning. In Proceedings of the 2008 conference on ECAI 2008: 18th European Conference on Artificial Intelligence. Amsterdam, The Netherlands, The Netherlands: IOS Press, pp. 234—238.
3. Možina, M. - Argument-based Machine Learning - http://www.ailab.si/martin/abml