ID2-of-3: Constructive Induction of M-of-N Concepts for Discriminators in Decision Trees

By P. Murphy, Michael J. Pazzani. 1991

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

A Survey of Methods for Explaining Black Box Models (2018, background): “ke into consideration only rules with conjunctions. Other types of rules are: m-of-n rules where given a set of nconditions if mof them are veried then the consequence of the rule is considered true [68]; list of rules where given an ordered set of rules is considered true the consequent of the rst ”
A historical perspective of explainable Artificial Intelligence (2020, methodology): “Craven and Shavlik (1995)'s approach can be seen as an extension of the ID2-of-3 algorithm (Murphy & Pazzani, 1991), a method for building decision trees from data based on “ m-of-n ” rules — that is, m out of n specified conditions must be true to send an example down a particular branch.”
The truth will come to light: directions and challenges in extracting the knowledge embedded within trained artificial neural networks (1998, result): “Analysis of results showed that TREPAN exhibits high fidelity with the model from which its trees are derived, exhibits high predictive accuracy, and is comparable in complexity with other decision trees trained on the same problem, e.g., C4.5 [30] and ID2-of-3 [ 25 ].”
ANN-DT: an algorithm for extraction of decision trees from artificial neural networks (1999, result): “Similar results were found by Craven and Shavlik [19], who compared the TREPAN algorithm to classification trees induced by C4.5 [3] and ID2-3 [ 34 ].”
Logistic Model Tree Extraction From Artificial Neural Networks (2007, background): “Trepan creates an n-of-m decision tree [32], which, in addition to the C4.”
Interpretable Rule Discovery Through Bilevel Optimization of Split-Rules of Nonlinear Decision Trees for Classification Problems (2020, methodology): “The study conducted in [14] uses a trained neural-network as an oracle to develop a DT of at least m-of-n type rule-sets (similar to the one described in ID2-of-3 [15]).”
AutoLearn — Automated Feature Generation and Selection (2017, background): “For example, it was shown that M-of-N expressions are particularly useful for medical classification problems where expert systems make use of criteria tables that are essentially M-of-N concepts [15].”
Feature Generation for Text Categorization Using World Knowledge (2005, unknown): “A number of feature generation algorithms were proposed [Pagallo and Haussler, 1990; Matheus and Rendell, 1989; Hu and Kibler, 1996; Murphy and Pazzani, 1991 ], which led to significant improvements in performance over a range of classification tasks.”
Extracting refined rules from knowledge-based neural networks (2004, methodology): “Further support for this method comes from experiments that indicate neural networks are good at learning M-of-N concepts (Fisher McKusick, 1989), as well as experiments with a variant of ID3 that show that a bias towards M-of-N style concepts is useful (Murphy & Pazzani, 1991).”
Trustworthy Artificial Intelligence: A Review (2022, methodology): “Another way to provide interpretability is through decision/rule sets [120], which use association rules like an if-then rule [120] or m-of-n rule [142] to generate classification rules.”