Fairness in Forecasting and Learning Linear Dynamical Systems

Authors: Quan Zhou, Jakub Marecek, Robert N. Shorten11134-11142

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

Reproducibility Variable Result LLM Response
Research Type Experimental Our empirical results on a biased data set motivated by insurance applications and the well-known COMPAS data set demonstrate both the beneficial impact of fairness considerations on statistical performance and the encouraging effects of exploiting sparsity on run time.
Researcher Affiliation Academia Quan Zhou,1 Jakub Marecek,2 Robert Shorten 1,3 1 University College Dublin, Ireland 2 Czech Technical University in Prague, the Czech Republic 3 Imperial College London, UK quan.zhou@ucdconnect.ie, jakub.marecek@fel.cvut.cz, r.shorten@imperial.ac.uk
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code Yes Our implementation is available on-line at https://github.com/Quan Zhou/Fairness-in-Learning-of-LDS.
Open Datasets Yes Our empirical results on a biased data set motivated by insurance applications and the well-known COMPAS data set demonstrate both the beneficial impact of fairness considerations on statistical performance and the encouraging effects of exploiting sparsity on run time. (Angwin et al. 2016)
Dataset Splits No The paper mentions 'training data' but does not specify explicit training/validation/test dataset splits (e.g., percentages, sample counts, or predefined splits) for reproducibility.
Hardware Specification Yes on a laptop equipped by Intel Core i7 8550U at 1.80 Ghz.
Software Dependencies No The paper mentions 'ncpol2sdpa of (Wittek 2015)' and 'sdpa of (Yamashita, Fujisawa, and Kojima 2003)' but does not provide specific version numbers for these software dependencies.
Experiment Setup Yes The three models (25)-(27) are applied in each experiment with λ of 1, 3, and 5, respectively, as chosen by iterating over integers 1 to 10. The initial states m(s) 0 of each subgroups are 5 and 7. We set the time window to be 20 across 3 trajectories in the advantaged subgroup and 2 in disadvantaged one, i.e., |T +| = 20, |I(a)| = 3 and |I(d)| = 2.