Of Neural Network Model Robustness Through Generating Invariant to Attributes Embeddings
Main Article Content
Abstract
Model robustness to minor deviations in the distribution of input data is an important criterion in many tasks. Neural networks show high accuracy on training samples, but the quality on test samples can be dropped dramatically due to different data distributions, a situation that is exacerbated at the subgroup level within each category. In this article we show how the robustness of the model at the subgroup level can be significantly improved with the help of the domain adaptation approach to image embeddings. We have found that application of a competitive approach to embeddings limitation gives a significant increase of accuracy metrics in a complex subgroup in comparison with the previous models. The method was tested on two independent datasets, the accuracy in a complex subgroup on the Waterbirds dataset is 90.3 {y : waterbirds;a : landbackground}, on the CelebA dataset is 92.22 {y : blondhair;a : male}.
Article Details
References
2. Christian Szegedy at all. Inception-v4, inception-resnet and the impact of residual connections on learning. Thirty-first AAAI Conference on Artificial Intelli-gence, 2017.
3. Dirk Hovy, Anders Søgaard. Tagging performance correlates with author age // Proceedings of the 53rd annual meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (volume 2: Short papers). 2015, P. 483–488.
4. Nicole Shadowen. Ethics and bias in machine learning: A technical study of what makes us “good”. The Transhumanism Handbook. Springer, 2019. P. 247–261.
5. Osonde A Osoba, William Welser IV. An intelligence in our image: The risks of bias and errors in artificial intelligence. Rand Corporation, 2017.
6. Shai Danziger, Jonathan Levavи, Liora Avnaim-Pesso. Extraneous factors in judicial decisions // Proceedings of the National Academy of Sciences 108.17 (2011). P. 6889–6892.
7. Amitabha Mukerjee at all. Multi-objective evolutionary algorithms for the risk-return trade-off in bank loan management // International Transactions in Operational Research 9.5. 2002. P. 583–597.
8. Julia K. Winkler at all. Association between surgical skin markings in der-moscopic images and diagnostic performance of a deep learning convolutional neural network for melanoma recognition // JAMA Dermatology 155.10. 2019. P. 1135–1141.
9. Philipp Tschandl, Cliff Rosendahl, Harald Kittler. The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions // Scientific Data 5. 2018. P. 180161.
10. Noel CF Codella at all. Skin lesion analysis toward melanoma detection: A challenge at the 2017 International Symposium on Biomedical Imaging (ISBI), hosted by the International Skin Imaging Collaboration (ISIC)// 2018 IEEE 15th International Symposium on Biomedical Imaging (ISBI 2018). IEEE. 2018. P. 168–172.
11. Marc Combalia at all. Bcn20000: Dermoscopic lesions in the wild // arXiv preprint arXiv:1908.02288, 2019.
12. Shiori Sagawa at all. Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization // arXiv preprint arXiv:1911.08731, 2019.
13. Sharon Li Karan Goel Albert Gu, Chris R´e. Automating the Art of Data Augmentation CLAMP: An Instantiation of Model Patching, 2020.
URL: http://hazyresearch.stanford.edu/data-aug-part-4.
14. Jun-Yan Zhu at all. Unpaired image-to-image translation using cyclecon-sistent adversarial networks // Proceedings of the IEEE international Conference on Computer Vision, 2017. P. 2223–2232.
15. Phillip Isola at all. Image-to-image translation with conditional adversari-al networks // Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017. P. 1125–1134.
16. Soumya Tripathy, Juho Kannala, Esa Rahtu. Learning image-to-image translation using paired and unpaired training samples // Asian Conference on Com-puter Vision. Springer, 2018. P. 51–66.
17. Ivan Anokhin at all. High-Resolution Daytime Translation Without Do-main Labels // arXiv preprint arXiv:2003.08791, 2020.
18. Tero Karras at all. Analyzing and improving the image quality of stylegan // arXiv preprint arXiv:1912.04958, 2019.
19. Sangwoo Mo, Minsu Cho, Jinwoo Shin. Instagan: Instance-aware imageto-image translation // arXiv preprint arXiv:1812.10889, 2018.
20. Yaroslav Ganin, Victor Lempitsky. Unsupervised domain adaptation by backpropagation // arXiv preprint arXiv:1409.7495, 2014.
21. Ying Tai, Jian Yang, Xiaoming Liu. Image super-resolution via deep recur-sive residual network // Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017. P. 3147–3155.
22. Jia Deng at all. Imagenet: A large-scale hierarchical image database// 2009 IEEE Conference on Computer Vision and Pattern Recognition. IEEE. 2009. P. 248–255.
23. Diederik P Kingma, Jimmy Ba. Adam: A method for stochastic optimiza-tion // arXiv preprint arXiv:1412.6980, 2014.
Presenting an article for publication in the Russian Digital Libraries Journal (RDLJ), the authors automatically give consent to grant a limited license to use the materials of the Kazan (Volga) Federal University (KFU) (of course, only if the article is accepted for publication). This means that KFU has the right to publish an article in the next issue of the journal (on the website or in printed form), as well as to reprint this article in the archives of RDLJ CDs or to include in a particular information system or database, produced by KFU.
All copyrighted materials are placed in RDLJ with the consent of the authors. In the event that any of the authors have objected to its publication of materials on this site, the material can be removed, subject to notification to the Editor in writing.
Documents published in RDLJ are protected by copyright and all rights are reserved by the authors. Authors independently monitor compliance with their rights to reproduce or translate their papers published in the journal. If the material is published in RDLJ, reprinted with permission by another publisher or translated into another language, a reference to the original publication.
By submitting an article for publication in RDLJ, authors should take into account that the publication on the Internet, on the one hand, provide unique opportunities for access to their content, but on the other hand, are a new form of information exchange in the global information society where authors and publishers is not always provided with protection against unauthorized copying or other use of materials protected by copyright.
RDLJ is copyrighted. When using materials from the log must indicate the URL: index.phtml page = elbib / rus / journal?. Any change, addition or editing of the author's text are not allowed. Copying individual fragments of articles from the journal is allowed for distribute, remix, adapt, and build upon article, even commercially, as long as they credit that article for the original creation.
Request for the right to reproduce or use any of the materials published in RDLJ should be addressed to the Editor-in-Chief A.M. Elizarov at the following address: amelizarov@gmail.com.
The publishers of RDLJ is not responsible for the view, set out in the published opinion articles.
We suggest the authors of articles downloaded from this page, sign it and send it to the journal publisher's address by e-mail scan copyright agreements on the transfer of non-exclusive rights to use the work.