Rule Generation from Neural Networks

By L. Fu. 1994

How researchers see this paper

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

Citations in the corpus

Explaining Explanations: An Overview of Interpretability of Machine Learning (2018, methodology): “The KT method [28] goes through each neuron, layer-by-layer and applies an if-then rule by ﬁnding a threshold.”
Rule Extraction Algorithm for Deep Neural Networks: A Review (2016, methodology): “Fu has proposed an algorithm called KT [19] that extract IF-THEN rule.”
Principles and Practice of Explainable Machine Learning (2020, background): “KT [53] is a related algorithm producing if-else rules, in a layer by layer manner.”
Inference is Everything: Recasting Semantic Resources into a Unified Evaluation Framework (2017, background): “Much prior work has aimed to improve the interpretability of neural networks, focusing in particular on extracting rules from the activations of feed forward networks (Towell and Shavlik, 1993; Thrun, 1993; Fu, 1994; Thrun, 1995).”
The truth will come to light: directions and challenges in extracting the knowledge embedded within trained artificial neural networks (1998, methodology): “Krishnan [22] provides a detailed analysis of the complexity of COMBO concluding that in the worst case COMBO, like other search based decompositional algorithms (Fu [11], [12]), is exponential.”
Extracting rules from multilayer perceptrons in classification problems: A clustering-based approach (2006, methodology): “Under this perspective, the Clustering Genetic Algorithm (CGA), whose main steps are summarized in Fig.”
Computational intelligence methods for rule-based data understanding (2004, methodology): “Local methods for extraction of conjunctive rules were proposed by Fu [68], [69], [70], [71], and Setiono and Liu [57], [58], [59], [72].”
Explaining Explanations: An Approach to Evaluating Interpretability of Machine Learning (2018, methodology): “The KT method [28] goes through each neuron, layer-by-layer and applies an if-then rule by ﬁnding a threshold.”
Extracting rules from trained neural networks (2000, background): “For example, [4] and [6] present the decompositional algorithms.”
Learning Global Additive Explanations for Neural Nets Using Model Distillation (2018, background): “, 2016; Frosst & Hinton, 2017) or model extraction (Fu, 1994; Sanchez et al., 2015; Lakkaraju et al., 2017; Bastani et al., 2017; Ribeiro et al., 2018).”
Interpreting individual classifications of hierarchical networks (2013, methodology): “Rule extraction refers to a family of methods that explain the general rules used by a network by analyzing the network’s internal weights [5].”
A Survey on the Explainability of Supervised Machine Learning (2020, background): “If a system avoids collisions sufficiently well, there are no further concerns about it. No explanation is required. Incompleteness can stem from various sources. Another example are safety reasons. For a system that cannot be tested in a full deployment environment, there is a certain incompletenes”
Methods and Procedures for the Verification and Validation of Artificial Neural Networks (2005, background): “The main assumption is that the recirculation network can perform back-propagation given a linear functional constrain for neurons in the visible layer, the weight matrix between the visible and hidden neurons is the symmetrical, and the visible neurons have a high regression. Hinton [1988] found th”
Extraction of rules from discrete-time recurrent neural networks (1996, background): “The extraction of (symbolic) rules from both feedforward as well as recurrent neural networks has become an active area of research (Cleeremans et al., 1989; Towell et al., 1990; Giles et al., 1992; Fu, 1994).”