Skip to main content
SpringerLink
Log in
Book cover

Joint European Conference on Machine Learning and Knowledge Discovery in Databases

ECML PKDD 2021: Machine Learning and Principles and Practice of Knowledge Discovery in Databases pp 90–96Cite as

Behavior of k-NN as an Instance-Based Explanation Method

  • Conference paper
  • First Online:

  • 2230 Accesses

Part of the book series: Communications in Computer and Information Science ((CCIS,volume 1524))

Abstract

Adoption of DL models in critical areas has led to an escalating demand for sound explanation methods. Instance-based explanation methods are a popular type that return selective instances from the training set to explain the predictions for a test sample. One way to connect these explanations with prediction is to ask the following counterfactual question - how does the loss and prediction for a test sample change when explanations are removed from the training set? Our paper answers this question for k-NNs which are natural contenders for an instance-based explanation method. We first demonstrate empirically that the representation space induced by last layer of a neural network is the best to perform k-NN in. Using this layer, we conduct our experiments and compare them to influence functions (IFs) [6] which try to answer a similar question. Our evaluations do indicate change in loss and predictions when explanations are removed but we do not find a trend between k and loss or prediction change. We find significant stability in the predictions and loss of MNIST vs. CIFAR-10. Surprisingly, we do not observe much difference in the behavior of k-NNs vs. IFs on this question. We attribute this to training set subsampling for IFs.

This is a preview of subscription content, log in via an institution.

Chapter
USD   29.95
Price excludes VAT (USA)
  • Available as PDF
  • Read on any device
  • Instant download
  • Own it forever
eBook
USD   129.00
Price excludes VAT (USA)
  • Available as EPUB and PDF
  • Read on any device
  • Instant download
  • Own it forever
Softcover Book
USD   169.99
Price excludes VAT (USA)
  • Compact, lightweight edition
  • Dispatched in 3 to 5 business days
  • Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Learn about institutional subscriptions

References

  1. Barshan, E., Brunet, M.E., Dziugaite, G.K.: RelatIF: identifying explanatory training samples via relative influence. In: International Conference on Artificial Intelligence and Statistics, pp. 1899–1909. PMLR (2020)

    Google Scholar 

  2. Caruana, R., Kangarloo, H., Dionisio, J.D., Sinha, U., Johnson, D.: Case-based explanation of non-case-based learning methods. In: Proceedings of the AMIA Symposium, p. 212. American Medical Informatics Association (1999)

    Google Scholar 

  3. Karimi, A.H., Barthe, G., Balle, B., Valera, I.: Model-agnostic counterfactual explanations for consequential decisions. In: International Conference on Artificial Intelligence and Statistics, pp. 895–905. PMLR (2020)

    Google Scholar 

  4. Kim, B., Koyejo, O., Khanna, R., et al.: Examples are not enough, learn to criticize! criticism for interpretability. In: NIPS, pp. 2280–2288 (2016)

    Google Scholar 

  5. Koh, P.W., Ang, K.S., Teo, H.H., Liang, P.: On the accuracy of influence functions for measuring group effects. arXiv preprint arXiv:1905.13289 (2019)

  6. Koh, P.W., Liang, P.: Understanding black-box predictions via influence functions. In: International Conference on Machine Learning, pp. 1885–1894. PMLR (2017)

    Google Scholar 

  7. Rajani, N.F., Krause, B., Yin, W., Niu, T., Socher, R., Xiong, C.: Explaining and improving model behavior with k nearest neighbor representations. arXiv preprint arXiv:2010.09030 (2020)

  8. Rousseeuw, L., Kaufman, P.: Clustering by means of medoids (1987)

    Google Scholar 

  9. Wachter, S., Mittelstadt, B., Russell, C.: Counterfactual explanations without opening the black box: automated decisions and the GDPR. Harv. JL & Tech. 31, 841 (2017)

    Google Scholar 

  10. Zagoruyko, S., Komodakis, N.: Wide residual networks. arXiv preprint arXiv:1605.07146 (2016)

Download references

Author information

Authors and Affiliations

  1. University of California, San Diego, USA

    Chhavi Yadav & Kamalika Chaudhuri

Authors
  1. Chhavi Yadav
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Kamalika Chaudhuri
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Chhavi Yadav .

Editor information

Editors and Affiliations

  1. IKIM, Ruhr-University Bochum, Bochum, Germany

    Michael Kamp

  2. University of Sydney, Sydney, NSW, Australia

    Irena Koprinska

  3. University of Namur, Namur, Belgium

    Adrien Bibal

  4. University of Rennes 1, Rennes, France

    Tassadit Bouadi

  5. University of Namur, Namur, Belgium

    Benoît Frénay

  6. Inria, Rennes, France

    Luis Galárraga

  7. University of Antwerp, Antwerp, Belgium

    José Oramas

  8. Ruhr University Bochum, Bochum, Germany

    Linara Adilova

  9. Royal Holloway University of London, Egham, UK

    Yamuna Krishnamurthy

  10. Ghent University, Ghent, Belgium

    Bo Kang

  11. Université Jean Monnet, Saint-Etienne cedex 2, France

    Christine Largeron

  12. Ghent University, Gent, Belgium

    Jefrey Lijffijt

  13. Telecom Paris, Paris, France

    Tiphaine Viard

  14. University of Bonn, Bonn, Germany

    Pascal Welke

  15. Norwegian Univesity of Science and Technology, Trondheim, Norway

    Massimiliano Ruocco

  16. BI Norwegian Business School, Oslo, Norway

    Erlend Aune

  17. University of Pisa, Pisa, Italy

    Claudio Gallicchio

  18. University of Duisburg-Essen, Essen, Germany

    Gregor Schiele

  19. Graz University of Technology, Graz, Austria

    Franz Pernkopf

  20. Xilinx Research, Dublin, Ireland

    Michaela Blott

  21. Heidelberg University, Heidelberg, Germany

    Holger Fröning

  22. Heidelberg University, Heidelberg, Germany

    Günther Schindler

  23. University of Pisa, Pisa, Italy

    Riccardo Guidotti

  24. University of Pisa, Pisa, Italy

    Anna Monreale

  25. ISTI-CNR, Pisa, Italy

    Salvatore Rinzivillo

  26. Warsaw University of Technology, Warsaw, Poland

    Przemyslaw Biecek

  27. Freie Universität Berlin, Berlin, Germany

    Eirini Ntoutsi

  28. Eindhoven University of Technology, Eindhoven, The Netherlands

    Mykola Pechenizkiy

  29. Leibniz University Hannover, Hannover, Germany

    Bodo Rosenhahn

  30. University of Sussex, Brighton, UK

    Christopher Buckley

  31. University of Chieti-Pescara, Chieti, Italy

    Daniela Cialfi

  32. Radboud University Nijmegen, Nijmegen, The Netherlands

    Pablo Lanillos

  33. McGill University, Montreal, Canada

    Maxwell Ramstead

  34. Ghent University, Ghent, Belgium

    Tim Verbelen

  35. University of Lisbon, Lisboa, Portugal

    Pedro M. Ferreira

  36. University of Bari Aldo Moro, Bari, Italy

    Giuseppina Andresini

  37. Universita di Bari Aldo Moro, Bari, Italy

    Donato Malerba

  38. University of Lisbon, Lisbon, Portugal

    Ibéria Medeiros

  39. Shenzhen University, Shenzhen, China

    Philippe Fournier-Viger

  40. Harbin Institute of Technology, Harbin, China

    M. Saqib Nawaz

  41. University of Córdoba, Córdoba, Spain

    Sebastian Ventura

  42. Peking University, Beijing, China

    Meng Sun

  43. Noah's Ark Lab, Huawei, Beijing, China

    Min Zhou

  44. UniCredit, Milan, Italy

    Valerio Bitetta

  45. UniCredit, Rome, Italy

    Ilaria Bordino

  46. UniCredit, Milan, Italy

    Andrea Ferretti

  47. Unicredit, Rome, Italy

    Francesco Gullo

  48. ENEA Headquarters, Portici, Italy

    Giovanni Ponti

  49. Unicredit, Rome, Italy

    Lorenzo Severini

  50. University of Porto, Porto, Portugal

    Rita Ribeiro

  51. University of Porto, Porto, Portugal

    João Gama

  52. UPC BarcelonaTech, Barcelona, Spain

    Ricard Gavaldà

  53. Northwestern University, Chicago, IL, USA

    Lee Cooper

  54. PD Personalised Healthcare, Basel, Switzerland

    Naghmeh Ghazaleh

  55. University of Lausanne, Lausanne, Switzerland

    Jonas Richiardi

  56. ETH Zurich, Basel, Switzerland

    Damian Roqueiro

  57. F. Hoffmann–La Roche Ltd, Basel, Switzerland

    Diego Saldana Miranda

  58. Novartis Pharma AG, Basel, Switzerland

    Konstantinos Sechidis

  59. University of Lisbon, Lisbon, Portugal

    Guilherme Graça

Rights and permissions

Reprints and permissions

Copyright information

© 2021 Springer Nature Switzerland AG

About this paper

Check for updates. Verify currency and authenticity via CrossMark

Cite this paper

Yadav, C., Chaudhuri, K. (2021). Behavior of k-NN as an Instance-Based Explanation Method. In: Kamp, M., et al. Machine Learning and Principles and Practice of Knowledge Discovery in Databases. ECML PKDD 2021. Communications in Computer and Information Science, vol 1524. Springer, Cham. https://doi.org/10.1007/978-3-030-93736-2_8

Download citation

  • DOI: https://doi.org/10.1007/978-3-030-93736-2_8

  • Published:

  • Publisher Name: Springer, Cham

  • Print ISBN: 978-3-030-93735-5

  • Online ISBN: 978-3-030-93736-2

  • eBook Packages: Computer ScienceComputer Science (R0)

Publish with us

Policies and ethics

Search

Navigation