Explanations for Monotonic Classifiers.

Authors: Joao Marques-Silva, Thomas Gerspacher, Martin C Cooper, Alexey Ignatiev, Nina Narodytska

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Section 4 summarizes initial experiments, which confirm the scalability of the proposed algorithms.
Researcher Affiliation Collaboration 1IRIT, CNRS, Universit e Paul Sabatier, Toulouse, France 2Monash University, Melbourne, Australia 3VMware Research, CA, USA.
Pseudocode Yes Algorithm 1 Finding one AXp find AXp(F, S, v); Algorithm 2 Finding one CXp find CXp(F, S, v); Algorithm 3 Enumeration of AXp s and CXp s
Open Source Code Yes XMono is available from https://git.io/JZZBX.
Open Datasets Yes COMET is run on the Auto-MPG11 dataset studied in earlier work (Sivaraman et al., 2020)... We use a monotonic subset (Pima Mono) of the Pima dataset12... The first dataset is Bankruptcy Risk (Greco et al., 1998)... 11http://tiny.cc/k3qytz. 12http://tiny.cc/l3qytz.
Dataset Splits No For each dataset, we either pick 100 instances, randomly selected, or the total number of instances in the dataset, in case this number does not exceed 100.
Hardware Specification Yes All experiments were run on a Mac Book Pro, with a 2.4GHz quad-core i5 processor, and 16 GByte of RAM, running Mac OS Big Sur.
Software Dependencies No All experiments were run on a Mac Book Pro, with a 2.4GHz quad-core i5 processor, and 16 GByte of RAM, running Mac OS Big Sur.
Experiment Setup No For each dataset, we either pick 100 instances, randomly selected, or the total number of instances in the dataset, in case this number does not exceed 100.