Does Multi-Clause Learning Help in Real-World Applications?

Abstract

Abstract. The ILP system Progol is incomplete in not being able to generalise a single example to multiple clauses. However, according to the Blumer bound, incomplete learners such as Progol, can have higher predictive accuracy using less search than more complete learners. This issue is particularly relevant in real-world problems, in which it is unclear whether the unknown target theory is within the hypothesis space of the incomplete learner. This paper uses two real-world applications in systems biology to study whether there exist datasets where a complete multi-clause learning (MCL) method can significantly outperform a single-clause learning (SCL) method. The experimental results show that in both applications there do exist datasets, in which hypotheses derived by MCL have significantly higher predictive accuracies. On the other hand, for most of the datasets in the two applications, there are good approximations of the target within the hypothesis space of SCL, so that MCL does not outperform SCL.

MC-TopLog: Complete Multi-clause Learning Guided by a Top Theory

“…That is why MCL is shown to have higher predictive accuracies than SCL later in the experiments of this paper. Such case also exists in other real-world applications, as demonstrated in [10].…”

“…This is further demonstrated in section 4 via experiments with two real-world data sets. More experiments with real-world applications can be found in [10], where target hypotheses are unknown for knowledge discovery tasks. The focus of this paper is to introduce a new complete approach called -directed theory coderivation( DTcD).…”

Construction and Validation of Food Webs Using Logic-Based Machine Learning and Text Mining

“…Modelling techniques can be broadly divided into three classes: Logical: These representations (Bernot et al, 2004;Calzone et al, 2006;Lin et al, 2012;Tamaddoni-Nezhad et al, 2007) are typically discrete, with particular strengths in the ease with which models can be understood by domain experts. Existing applications include shape-oriented models of large and small molecules, as well as biochemical network representations.…”

“…The area is divided into various sub-topics that are largely related to the form of computational model employed, many of which are presently being applied to complex biological systems (Bernot et al, 2004;Calzone et al, 2006;Chen and Xu, 2004;Dale et al, 2010;Lin et al, 2012;Mazandu and Mulder, 2012;Tamaddoni-Nezhad et al, 2007;Xiong et al, 2006).…”