An evolutionary system for neural logic networks using genetic programming and indirect encoding

Abstract

Nowadays, intelligent connectionist systems such as artificial neural networks have been proved very powerful in a wide area of applications. Consequently, the ability to interpret their structure was always a desirable feature for experts. In this field, the neural logic networks (NLN) by their definition are able to represent complex human logic and provide knowledge discovery. However, under contemporary methodologies, the training of these networks may often result in non-comprehensible or poorly designed structures. In this work, we propose an evolutionary system that uses current advances in genetic programming that overcome these drawbacks and produces neural logic networks that can be arbitrarily connected and are easily interpretable into expert rules. To accomplish this task, we guide the genetic programming process using a context-free grammar and we encode indirectly the neural logic networks into the genetic programming individuals. We test the proposed system in two problems of medical diagnosis. Our results are examined both in terms of the solution interpretability that can lead in knowledge discovery, and in terms of the achieved accuracy. We draw conclusions about the effectiveness of the system and we propose further research directions.

Breast tumor malignancy modelling using evolutionary neural logic networks

“…In the full data that are now available for experimentation (699 records), a subset of 458 (65.5%) concern benign tumors and the rest 241 (34.5%) correspond to malignant ones. The database features are integer numbers in [1,10] corresponding to a quantitative characterization of nine laboratory measurements (T1-T9) of the cells, presented in detail in Table I. To avoid overfitting during the training phase of our system, we made use of a validation set.…”

“…Specifically, the evolutionary methodology guides the formation and the tuning of a neural computation model, constructing finally an evolutionary neural logic network (ENLN), described in detail previously (1). An advanced evolutionary computation approach is used, namely grammar-guided genetic programming (GGGP), using cellular encoding.…”

Free lunches for neural network search

“…There is an enormous amount of literature on the application of evolutionary algorithms (EAs) to the synthesis of artificial neural networks (NNs), this topic having been very popular for over two decades [14,24,9,2,3,29,4,18,15,19,11,22,1,7]. Research in this area, which we will term Evolutionary Neural Networks (ENNs) hereafter, can be divided in three branches.…”

Towards neural-symbolic integration: the evolutionary neural logic networks

“…Therefore, when the syntax form of the desired solution is already known, it is (4,4) Car_Owner (Χ,Υ)…”

“…Moreover, a solution obtained this way, leads to potential knowledge extraction. Very recently, a new system, namely the evolutionary neural logic networks (ENLN), has been proposed [4] that fulfills those requirements. The new approach uses grammar-guided genetic programming to produce neural logic networks.…”