Feature and Label Embedding Spaces Matter in Addressing Image Classifier Bias

This paper strives to address image classifier bias, with a focus on both feature and label embedding spaces. Previous works have shown that spurious correlations from protected attributes, such as age, gender, or skin tone, can cause adverse decisions. To balance potential harms, there is a growing need to identify and mitigate image classifier bias. First, we identify in the feature space a bias direction. We compute class prototypes of each protected attribute value for every class, and reveal an existing subspace that captures the maximum variance of the bias. Second, we mitigate biases by mapping image inputs to label embedding spaces. Each value of the protected attribute has its projection head where classes are embedded through a latent vector representation rather than a common one-hot encoding. Once trained, we further reduce in the feature space the bias effect by removing its direction. Evaluation on biased image datasets, for multi-class, multi-label and binary classifications, shows the effectiveness of tackling both feature and label embedding spaces in improving the fairness of the classifier predictions, while preserving classification performance.

[1]  Sixue Gong,et al.  Jointly De-Biasing Face Recognition and Demographic Attribute Estimation , 2019, ECCV.

[2]  Jeffrey Dean,et al.  Efficient Estimation of Word Representations in Vector Space , 2013, ICLR.

[3]  Michael S. Bernstein,et al.  ImageNet Large Scale Visual Recognition Challenge , 2014, International Journal of Computer Vision.

[4]  Jian Sun,et al.  Deep Residual Learning for Image Recognition , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[5]  Trevor Darrell,et al.  Simultaneous Deep Transfer Across Domains and Tasks , 2015, 2015 IEEE International Conference on Computer Vision (ICCV).

[6]  Nathan Srebro,et al.  Equality of Opportunity in Supervised Learning , 2016, NIPS.

[7]  Zhe Zhao,et al.  Data Decisions and Theoretical Implications when Adversarially Learning Fair Representations , 2017, ArXiv.

[8]  François Laviolette,et al.  Domain-Adversarial Training of Neural Networks , 2015, J. Mach. Learn. Res..

[9]  Olga Russakovsky,et al.  Towards Fairness in Visual Recognition: Effective Strategies for Bias Mitigation , 2019, 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

[10]  Laurens van der Maaten,et al.  Does Object Recognition Work for Everyone? , 2019, CVPR Workshops.

[11]  Timnit Gebru,et al.  Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification , 2018, FAT.

[12]  Mayank Vatsa,et al.  On the Robustness of Face Recognition Algorithms Against Attacks and Bias , 2020, AAAI.

[13]  Arvind Narayanan,et al.  Mitigating dataset harms requires stewardship: Lessons from 1000 papers , 2021, NeurIPS Datasets and Benchmarks.

[14]  M. Bethge,et al.  Shortcut learning in deep neural networks , 2020, Nature Machine Intelligence.

[15]  Vinay Uday Prabhu,et al.  Large image datasets: A pyrrhic win for computer vision? , 2020, 2021 IEEE Winter Conference on Applications of Computer Vision (WACV).

[16]  Rachel K. E. Bellamy,et al.  AI Fairness 360: An Extensible Toolkit for Detecting, Understanding, and Mitigating Unwanted Algorithmic Bias , 2018, ArXiv.

[17]  Junmo Kim,et al.  Learning Not to Learn: Training Deep Neural Networks With Biased Data , 2018, 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

[18]  Sean A. Munson,et al.  Unequal Representation and Gender Stereotypes in Image Search Results for Occupations , 2015, CHI.

[19]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[20]  H. Byun,et al.  Exploiting Transferable Knowledge for Fairness-Aware Image Classification , 2020, ACCV.

[21]  Alexei A. Efros,et al.  Unbiased look at dataset bias , 2011, CVPR 2011.

[22]  Luke S. Zettlemoyer,et al.  Deep Contextualized Word Representations , 2018, NAACL.

[23]  Adam Tauman Kalai,et al.  Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings , 2016, NIPS.

[24]  Alex Krizhevsky,et al.  Learning Multiple Layers of Features from Tiny Images , 2009 .

[25]  Oluwasanmi Koyejo,et al.  Examples are not enough, learn to criticize! Criticism for Interpretability , 2016, NIPS.

[26]  Arvind Narayanan,et al.  Semantics derived automatically from language corpora contain human-like biases , 2016, Science.

[27]  Emily Denton,et al.  Towards Accountability for Machine Learning Datasets: Practices from Software Engineering and Infrastructure , 2020, FAccT.

[28]  Pieter Abbeel,et al.  InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets , 2016, NIPS.

[29]  Ryan Cotterell,et al.  Gender Bias in Contextualized Word Embeddings , 2019, NAACL.

[30]  Daniel Jurafsky,et al.  Word embeddings quantify 100 years of gender and ethnic stereotypes , 2017, Proceedings of the National Academy of Sciences.

[31]  Timnit Gebru,et al.  Datasheets for datasets , 2018, Commun. ACM.

[32]  Yoav Goldberg,et al.  Lipstick on a Pig: Debiasing Methods Cover up Systematic Gender Biases in Word Embeddings But do not Remove Them , 2019, NAACL-HLT.

[33]  Geoffrey E. Hinton,et al.  On the importance of initialization and momentum in deep learning , 2013, ICML.

[34]  Jieyu Zhao,et al.  Balanced Datasets Are Not Enough: Estimating and Mitigating Gender Bias in Deep Image Representations , 2018, 2019 IEEE/CVF International Conference on Computer Vision (ICCV).

[35]  Fei-Fei Li,et al.  Towards fairer datasets: filtering and balancing the distribution of the people subtree in the ImageNet hierarchy , 2019, FAT*.

[36]  Kimmo Kärkkäinen,et al.  FairFace: Face Attribute Dataset for Balanced Race, Gender, and Age , 2019, ArXiv.

[37]  Luc Van Gool,et al.  Deep Expectation of Real and Apparent Age from a Single Image Without Facial Landmarks , 2016, International Journal of Computer Vision.

[38]  Yi Yang,et al.  Taking a Closer Look at Domain Shift: Category-Level Adversaries for Semantics Consistent Domain Adaptation , 2018, 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

[39]  Tatsuya Harada,et al.  Maximum Classifier Discrepancy for Unsupervised Domain Adaptation , 2017, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[40]  Toniann Pitassi,et al.  Fairness through awareness , 2011, ITCS '12.

[41]  Lucy Vasserman,et al.  Measuring and Mitigating Unintended Bias in Text Classification , 2018, AIES.

[42]  Moustapha Cissé,et al.  ConvNets and ImageNet Beyond Accuracy: Understanding Mistakes and Uncovering Biases , 2017, ECCV.

[43]  Olga Russakovsky,et al.  REVISE: A Tool for Measuring and Mitigating Bias in Visual Datasets , 2020, International Journal of Computer Vision.

[44]  D. Sculley,et al.  No Classification without Representation: Assessing Geodiversity Issues in Open Data Sets for the Developing World , 2017, 1711.08536.

[45]  Andrew Zisserman,et al.  Turning a Blind Eye: Explicit Removal of Biases and Variation from Deep Neural Network Embeddings , 2018, ECCV Workshops.

[46]  Yoshua Bengio,et al.  Generative Adversarial Nets , 2014, NIPS.

[47]  Emily M. Bender,et al.  On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜 , 2021, FAccT.

[48]  Yuxiao Hu,et al.  MS-Celeb-1M: A Dataset and Benchmark for Large-Scale Face Recognition , 2016, ECCV.

[49]  Blake Lemoine,et al.  Mitigating Unwanted Biases with Adversarial Learning , 2018, AIES.

[50]  Mei Wang,et al.  Racial Faces in the Wild: Reducing Racial Bias by Information Maximization Adaptation Network , 2018, 2019 IEEE/CVF International Conference on Computer Vision (ICCV).

[51]  Trevor Darrell,et al.  Women also Snowboard: Overcoming Bias in Captioning Models , 2018, ECCV.

[52]  Xiaogang Wang,et al.  Deep Learning Face Attributes in the Wild , 2014, 2015 IEEE International Conference on Computer Vision (ICCV).

[53]  Jieyu Zhao,et al.  Men Also Like Shopping: Reducing Gender Bias Amplification using Corpus-level Constraints , 2017, EMNLP.

[54]  Xiang Yu,et al.  Feature Transfer Learning for Face Recognition With Under-Represented Data , 2019, 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).