Satisfying Real-world Goals with Dataset Constraints

Authors: Gabriel Goh, Andrew Cotter, Maya Gupta, Michael P. Friedlander

NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on both benchmark and real-world industry datasets demonstrate the effectiveness of our approach.
Researcher Affiliation Collaboration Gabriel Goh Dept. of Mathematics UC Davis Davis, CA 95616 ggoh@math.ucdavis.edu Andrew Cotter, Maya Gupta Google Inc. 1600 Amphitheatre Parkway Mountain View, CA 94043 acotter@google.com mayagupta@google.com Michael Friedlander Dept. of Computer Science University of British Columbia Vancouver, B.C. V6T 1Z4 mpf@cs.ubc.ca
Pseudocode Yes Algorithm 1 Proposed majorization-minimization procedure for (approximately) optimizing Problem 2... Algorithm 2 Skeleton of a cutting-plane algorithm that optimizes Equation 6...
Open Source Code Yes Our publicly-available Julia implementation3 for these experiments uses LIBLINEAR [11] with the default parameters... 3https://github.com/gabgoh/svmc.jl
Open Datasets Yes We compare training for fairness on the Adult dataset 2, the same dataset used by Zafar et al. [27]. ... 2 a9a from https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary.html
Dataset Splits Yes The 32 561 training and 16 281 testing examples, derived from the 1994 Census... For all three datasets, we split out 80% for training and reserved 20% for testing.
Hardware Specification No The paper does not provide specific details about the hardware used for running its experiments.
Software Dependencies No The paper mentions using a 'Julia implementation' and 'LIBLINEAR' but does not specify their version numbers in the main text.
Experiment Setup Yes Our publicly-available Julia implementation3 for these experiments uses LIBLINEAR [11] with the default parameters (most notably λ = 1/n 3 10 5) to implement the SVMOptimizer function, and does not include an unregularized bias b. ... We performed 5 iterations of the majorization-minimization procedure of Algorithm 1. ... We chose the regularization parameter λ using a power-of-10 grid search, found that 10 7 was best for this baseline, and then used λ = 10 7 for all experiments.