Towards Threshold Invariant Fair Classification

被引:0
|
作者
Chen, Mingliang [1 ]
Wu, Min
机构
[1] Univ Maryland, Elect & Comp Engn Dept, College Pk, MD 20742 USA
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Effective machine learning models can automatically learn useful information from a large quantity of data and provide decisions in a high accuracy. These models may, however, lead to unfair predictions in certain sense among the population groups of interest, where the grouping is based on such sensitive attributes as race and gender. Various fairness definitions, such as demographic parity and equalized odds, were proposed in prior art to ensure that decisions guided by the machine learning models are equitable. Unfortunately, the "fair" model trained with these fairness definitions is threshold sensitive, i.e., the condition of fairness may no longer hold true when tuning the decision threshold. This paper introduces the notion of threshold invariant fairness, which enforces equitable performances across different groups independent of the decision threshold. To achieve this goal, this paper proposes to equalize the risk distributions among the groups via two approximation methods. Experimental results demonstrate that the proposed methodology is effective to alleviate the threshold sensitivity in machine learning models designed to achieve fairness.
引用
收藏
页码:560 / 569
页数:10
相关论文
共 50 条
  • [1] Towards Fair and Robust Classification
    Sun, Haipei
    Wu, Kun
    Wang, Ting
    Wang, Wendy Hui
    [J]. 2022 IEEE 7TH EUROPEAN SYMPOSIUM ON SECURITY AND PRIVACY (EUROS&P 2022), 2022, : 356 - 376
  • [2] Group-Aware Threshold Adaptation for Fair Classification
    Jang, Taeuk
    Shi, Pengyi
    Wang, Xiaoqian
    [J]. THIRTY-SIXTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE / THIRTY-FOURTH CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE / THE TWELVETH SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2022, : 6988 - 6995
  • [3] FAIR:: Towards a new feature for affinely-invariant recognition
    Sara, Radim
    Matousek, Martin
    [J]. 18TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION, VOL 2, PROCEEDINGS, 2006, : 412 - +
  • [4] Towards an universal classification of scale invariant processes
    B. Dubrulle
    F.-M. Bréon
    F. Graner
    A. Pocheau
    [J]. The European Physical Journal B - Condensed Matter and Complex Systems, 1998, 4 : 89 - 94
  • [5] Conformally invariant quantization - towards the complete classification
    Silhan, Josef
    [J]. DIFFERENTIAL GEOMETRY AND ITS APPLICATIONS, 2014, 33 : 162 - 176
  • [6] Towards an universal classification of scale invariant processes
    Dubrulle, B
    Breon, FM
    Graner, F
    Pocheau, A
    [J]. EUROPEAN PHYSICAL JOURNAL B, 1998, 4 (01): : 89 - 94
  • [7] Adversarial Debiasing techniques towards 'fair' skin lesion classification
    Correa-Medero, Ramon L.
    Patel, Bhavik
    Banerjee, Imon
    [J]. 2023 11TH INTERNATIONAL IEEE/EMBS CONFERENCE ON NEURAL ENGINEERING, NER, 2023,
  • [8] Towards Robust Classification Model by Counterfactual and Invariant Data Generation
    Chang, Chun-Hao
    Adam, George Alexandru
    Goldenberg, Anna
    [J]. 2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021, 2021, : 15207 - 15216
  • [9] Towards threshold-agnostic heavy-hitter classification
    Pekar, Adrian
    Duque-Torres, Alejandra
    Seah, Winston K. G.
    Caicedo Rendon, Oscar M.
    [J]. INTERNATIONAL JOURNAL OF NETWORK MANAGEMENT, 2022, 32 (03)
  • [10] Towards pose-invariant 2D face classification for surveillance
    Sanderson, Conrad
    Shang, Ting
    Lovell, Brian C.
    [J]. ANALYSIS AND MODELING OF FACES AND GESTURES, PROCEEDINGS, 2007, 4778 : 276 - 289