An Efficient Pruning Algorithm for Robust Isotonic Regression

Authors: Cong Han Lim

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments show that this algorithm can perform much faster than the dynamic programming approach on robust estimators, especially as the desired accuracy increases.
Researcher Affiliation Academia Cong Han Lim School of Industrial Systems and Engineering Georgia Tech Altanta, GA 30332 clim31@gatech.edu Work done while at Wisconsin Institute for Discovery, University of Wisconsin-Madison.
Pseudocode Yes Algorithm 1 Dynamic Program for fixed grid Gk; Algorithm 2 Algorithmic Framework for Faster Robust Isotonic Regression; Algorithm 3 Pruning I via Lower/Upper Bounds; Algorithm 4 Pruning Subroutine; Algorithm 5 Main Algorithm for Refining via Linear Underestimators; Algorithm 6 A Pruning Algorithm for Robust Isotonic Regression
Open Source Code No The paper does not provide any explicit statements about making the source code available or provide any links to a code repository.
Open Datasets No The paper describes generating synthetic data: 'We generate a series of n points y1, . . . , yn from 0.2 to 0.8 equally spaced out and added Gaussian random noise with standard deviation of 0.03. We then randomly flipped between 5% to 50% of the points around 0.5, and these points act as the outliers.' It does not provide access information for a publicly available or open dataset.
Dataset Splits No The paper describes how the data was generated and used in experiments but does not specify any training, validation, or test dataset splits.
Hardware Specification Yes The algorithms were implemented in Python 3.6.7, and the experiments were ran on an Intel 7th generation core i5-7200U dual-core processor with 8GB of RAM.
Software Dependencies No The paper states: 'The algorithms were implemented in Python 3.6.7'. It does not list specific libraries or specialized solvers with version numbers.
Experiment Setup Yes We set n to 1000 and varied k from 27 = 128 to 216 = 65536. We then randomly flipped between 5% to 50% of the points around 0.5, and these points act as the outliers. The results are averaged over 10 independent trials.