Rule Extraction Algorithm for Deep Neural Networks: A Review

Knowledge and Information Systems, 2020

Representing the knowledge learned by neural networks in the form of interpretable rules is a prudent technique to justify the decisions made by neural networks. Heretofore many algorithms exist to extract symbolic rules from neural networks, but among them, a few extract rules from deep neural networks trained using deep learning techniques. So, this paper proposes an algorithm to extract rules from a multi-hidden layer neural network, pre-trained using deep belief network and fine-tuned using back propagation. The algorithm analyzes each node of a layer and extracts knowledge from each layer separately. The process of knowledge extraction from the first hidden layer is different from the other layers. Consecutively, the algorithm combines all the knowledge extracted and refines them to construct a final ruleset consisting of symbolic rules. The algorithm further subdivides the subspace of a rule in the ruleset if it satisfies certain conditions. Results show that the algorithm extracted rules with higher accuracy compared to some existing rule extraction algorithms. Other than accuracy, the efficacy of the extracted rules is also validated with fidelity and various other performance measures.

This paper presents a critical review ofthe rule extraction techniques used for classification. The aspects of ability of the different rule extraction methods to solve classification problem are enlightened. In particular, special attention is given to rule extraction literature related with learning algorithms. The extensive review of Rule extraction in neural network will provide necessary guidance to the critical review comments that identify some gap in the rule extraction techniques in related literature

International Conference on Pattern Recognition, Informatics and Medical Engineering (PRIME-2012), 2012

Though neural networks have achieved highest classification accuracy for many classification problems, the obtained results may not be interpretable as they are often considered as black box. To overcome this drawback researchers have developed many rule extraction algorithms. This paper has discussed on various rule extraction algorithms based on three different rule extraction approaches namely decompositional, pedagogical and eclectic. Also it evaluates the performance of those approaches by comparing different algorithms with these three approaches on three real datasets namely Wisconsin breast cancer, Pima Indian diabetes and Iris plants.

2016

Classification and Rule extraction is an important application of Artificial Neural Network. To extract fewer rules from multilayer feed forward neural network has been a research area. The internal representation of the network is augmented by a distance term to extract fewer rules from the feed forward neural network and experimented on five datasets. Understanding affect of different factors of the dataset and network on extraction of a number of rules from the network can reveal important pieces of information which may help researchers to enhance the rule extraction process. This work investigates the internal behavior of neural network in rule extraction process on five different dataset. Keywords: Rule extraction, Feed Forward Neural Network, Hidden units, Activation value, Hidden neurons.

International Journal of Futuristic Trends in Engineering and Technology(IJFTET)(ISSN 2348 - 4071), 2014

Abstract: Although neural networks have performed very well for many application domains, one of its main drawbacks is the inherent black-box nature. Many researchers have developed techniques to extract human understandable symbolic rules from neural network. Here the survey of these techniques according to their approaches is presented. The Paper focuses on the basics of neural network and the Approaches and Techniques of extracting classification Rules from Neural Network. Comparison of approaches is also present.

2021

This article presents the basic methods of machine learning and explanational artificial intelligence that can help in the issue of extracting rules and other models of knowledge representation not only from data, but from the artificial neural networks themselves. The paper discusses classification methods for rule-based learning methods for neural networks and the current state of technologies for extracting rules from neural networks. Next, we formulate the main problems that arise when extracting rules from neural networks, as well as the main methods for solving them.

New Generation Computing, 2018

Neural network is one of the best tools for data mining tasks due to its high accuracy. However, one of the drawbacks of neural network is its black box nature. This limitation makes neural network useless for many applications which require transparency in their decision-making process. Many algorithms have been proposed to overcome this drawback by extracting transparent rules from neural network, but still researchers are in search for algorithms that can generate more accurate and simple rules. Therefore, this paper proposes a rule extraction algorithm named Eclectic Rule Extraction from Neural Network Recursively (ERENNR), with the aim to generate simple and accurate rules. ERENNR algorithm extracts symbolic classification rules from a single-layer feed-forward neural network. The novelty of this algorithm lies in its procedure of analyzing the nodes of the network. It analyzes a hidden node based on data ranges of input attributes with respect to its output and analyzes an output node using logical combination of the outputs of hidden nodes with respect to output class. And finally it generates a rule set by proceeding in a backward direction starting from the output layer. For each rule in the set, it repeats the whole process of rule extraction if the rule satisfies certain criteria. The algorithm is validated with eleven benchmark datasets. Experimental results show that the generated rules are simple and accurate.

Lecture Notes in Computer Science, 2018

Convolutional Neural Networks (CNNs) lack an explanation capability in the form of propositional rules. In this work we define a simple CNN architecture having a unique convolutional layer, then a Max-Pool layer followed by a full connected layer. Rule extraction is performed after the Max-Pool layer with the use of the Discretized Interpretable Multi Layer Perceptron (DIMLP). The antecedents of the extracted rules represent responses of convolutional filters, which are difficult to understand. However, we show in a sentiment analysis problem that from these "meaningless" values it is possible to obtain rules that represent relevant words in the antecedents. The experiments illustrate several examples of rules that represent n-grams.

International journal of neural systems, 2001

The problem of rule extraction from neural networks is NP-hard. This work presents a new technique to extract "if-then-else" rules from ensembles of DIMLP neural networks. Rules are extracted in polynomial time with respect to the dimensionality of the problem, the number of examples, and the size of the resulting network. Further, the degree of matching between extracted rules and neural network responses is 100%. Ensembles of DIMLP networks were trained on four data sets in the public domain. Extracted rules were on average significantly more accurate than those extracted from C4.5 decision trees.

Applied Soft Computing, 2010

Artificial neural network (ANN) is one of the most widely used techniques in classification data mining. Although ANNs can achieve very high classification accuracies, their explanation capability is very limited. Therefore one of the main challenges in using ANNs in data mining applications is to extract explicit knowledge from them. Based on this motivation, a novel approach is proposed in this paper for generating classification rules from feed forward type ANNs. Although there are several approaches in the literature for classification rule extraction from ANNs, the present approach is fundamentally different from them. In the previous studies, ANN training and rule extraction is generally performed independently in a sequential (hierarchical) manner. However, in the present study, training and rule extraction phases are integrated within a multiple objective evaluation framework for generating accurate classification rules directly. The proposed approach makes use of differential evolution algorithm for training and touring ant colony optimization algorithm for rule extracting. The proposed algorithm is named as DIFACONN-miner. Experimental study on the benchmark data sets and comparisons with some other classical and state-of-the art rule extraction algorithms has shown that the proposed approach has a big potential to discover more accurate and concise classification rules.