Extracting refined rules from knowledge-based neural networks

By Geoffrey G. Towell, Jude W. Shavlik. 2004

297 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Safety, Robustness & Interpretability, Knowledge Graphs & Representation

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Safety, Robustness & Interpretability

In Knowledge Graphs & Representation

Citations in the corpus

Rule Extraction Algorithm for Deep Neural Networks: A Review (2016, methodology): “The rule has the following form: This method is efficient and general [15].”
Supersparse linear integer models for optimized medical scoring systems (2015, methodology): “M-of-N rule tables were originally proposed as auxiliary models that could be extracted from neural nets (Towell and Shavlik, 1993) but can also be trained as stand-alone discrete linear classification models as suggested by Chevaleyre et al. (2013).”
Or's of And's for Interpretable Classification, with Application to Context-Aware Recommender Systems (2015, methodology): “Some of these methods, like CBA, CPAR and CMAR [10, 11, 26, 51] still suffer from a huge number of patterns and do not yield interpretable classiﬁers, yet it is well-known that for many domains, the space of good predictive models is often large enough to include very simple models [20].”
Methods and Models for Interpretable Linear Classification (2014, background): “M-of-N rule tables were originally proposed as auxiliary models that could be extracted from neural nets [108].”
Explaining inferences in Bayesian networks (2008, methodology): “The EBI procedure uses decision tree rules to represent variable interactions, but alternative forms of representation like M-of-N rules [34, 35] may provide a richer semantics.”
Extracting rules from multilayer perceptrons in classification problems: A clustering-based approach (2006, methodology): “The rule extraction process is visualized as a learning task, whose main goal is to learn the function computed by the trained NN, providing conjunctive and MofN [68] rules.”
Noisy Time Series Prediction using Recurrent Neural Networks and Grammatical Inference (2001, background): “A number of people have considered the extraction of symbolic knowledge from trained neural networks for both feedforward and recurrent neural networks [12, 44, 57, 44].”
Efficient Filtering on Hidden Document Streams (2014, background): “" Many other types of classifiers like decision trees (Quinlan 1987) and linear classifiers (Towell and Shavlik 1993; Fung, Sandilya, and Rao 2005) can be easily converted to sets of rules; hence we believe that our modeling is quite general.”
Combining Symbolic and Neural Learning (1994, methodology): “Towell and Shavlik (1993) developed a method that produces about one "N out of M" rule for each node.”
Orthogonal search-based rule extraction (OSRE) for trained neural networks: a practical and efficient approach (2006, background): “Examples of decompositional rule extraction algorithms are found in [5], [6], [7] & [8].”
An Efficient Explanation of Individual Classifications using Game Theory (2010, background): “The explanation and interpretation of artificial neural networks, which are arguably one of the least transparent models, has also received a lot of attention, especially in the form of rule extraction (Towell and Shavlik, 1993; Andrews et al., 1995; Nayak, 2009).”
Rule extraction from trained adaptive neural networks using artificial immune systems (2009, methodology): “Towell and Shavlik (1993) showed how to use ANNs for rule refinement.”