Advertisement

Explaining and Interpreting LSTMs

  • Leila Arras
  • José Arjona-Medina
  • Michael Widrich
  • Grégoire Montavon
  • Michael Gillhofer
  • Klaus-Robert Müller
  • Sepp Hochreiter
  • Wojciech SamekEmail author
Chapter
Part of the Lecture Notes in Computer Science book series (LNCS, volume 11700)

Abstract

While neural networks have acted as a strong unifying force in the design of modern AI systems, the neural network architectures themselves remain highly heterogeneous due to the variety of tasks to be solved. In this chapter, we explore how to adapt the Layer-wise Relevance Propagation (LRP) technique used for explaining the predictions of feed-forward networks to the LSTM architecture used for sequential data modeling and forecasting. The special accumulators and gated interactions present in the LSTM require both a new propagation scheme and an extension of the underlying theoretical framework to deliver faithful explanations.

Keywords

Explainable artificial intelligence Model transparency Recurrent neural networks LSTM Interpretability 

Notes

Acknowledgements

This work was supported by the German Ministry for Education and Research as Berlin Big Data Centre (01IS14013A), Berlin Center for Machine Learning (01IS18037I) and TraMeExCo (01IS18056A). Partial funding by DFG is acknowledged (EXC 2046/1, project-ID: 390685689). This work was also supported by the Institute for Information & Communications Technology Planning & Evaluation (IITP) grant funded by the Korea government (No. 2017-0-00451, No. 2017-0-01779).

References

  1. 1.
    Ancona, M., Ceolini, E., Öztireli, C., Gross, M.: Towards better understanding of gradient-based attribution methods for deep neural networks. In: International Conference on Learning Representations (ICLR) (2018)Google Scholar
  2. 2.
    Arjona-Medina, J.A., Gillhofer, M., Widrich, M., Unterthiner, T., Brandstetter, J., Hochreiter, S.: RUDDER: return decomposition for delayed rewards. arXiv:1806.07857 (2018)
  3. 3.
    Arras, L., Horn, F., Montavon, G., Müller, K.R., Samek, W.: “What is relevant in a text document?”: An interpretable machine learning approach. PLoS ONE 12(8), e0181142 (2017)CrossRefGoogle Scholar
  4. 4.
    Arras, L., Montavon, G., Müller, K.R., Samek, W.: Explaining recurrent neural network predictions in sentiment analysis. In: Proceedings of the EMNLP 2017 Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis (WASSA), pp. 159–168 (2017)Google Scholar
  5. 5.
    Arras, L., Osman, A., Müller, K.R., Samek, W.: Evaluating recurrent neural network explanations. In: Proceedings of the ACL 2019 Workshop on BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, pp. 113–126. Association for Computational Linguistics (2019)Google Scholar
  6. 6.
    Bach, S., Binder, A., Montavon, G., Klauschen, F., Müller, K.R., Samek, W.: On pixel-wise explanations for non-linear classifier decisions by layer-wise relevance propagation. PLoS ONE 10(7), e0130140 (2015)CrossRefGoogle Scholar
  7. 7.
    Bakker, B.: Reinforcement learning with long short-term memory. In: Advances in Neural Information Processing Systems 14 (NIPS), pp. 1475–1482 (2002)Google Scholar
  8. 8.
    Bakker, B.: Reinforcement learning by backpropagation through an LSTM model/ critic. In: IEEE International Symposium on Approximate Dynamic Programming and Reinforcement Learning, pp. 127–134 (2007)Google Scholar
  9. 9.
    Becker, S., Ackermann, M., Lapuschkin, S., Müller, K.R., Samek, W.: Interpreting and explaining deep neural networks for classification of audio signals. arXiv:1807.03418 (2018)
  10. 10.
    Bengio, Y., Simard, P., Frasconi, P.: Learning long-term dependencies with gradient descent is difficult. IEEE Trans. Neural Networks 5(2), 157–166 (1994)CrossRefGoogle Scholar
  11. 11.
    Chen, J., Song, L., Wainwright, M., Jordan, M.: Learning to explain: an information-theoretic perspective on model interpretation. In: Proceedings of the 35th International Conference on Machine Learning (ICML), vol. 80, pp. 883–892 (2018)Google Scholar
  12. 12.
    Cho, K., et al.: Learning phrase representations using RNN encoder-decoder for statistical machine translation. In: Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp. 1724–1734. Association for Computational Linguistics (2014)Google Scholar
  13. 13.
    Denil, M., Demiraj, A., de Freitas, N.: Extraction of salient sentences from labelled documents. arXiv:1412.6815 (2015)
  14. 14.
    Ding, Y., Liu, Y., Luan, H., Sun, M.: Visualizing and understanding neural machine translation. In: Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (ACL), pp. 1150–1159. Association for Computational Linguistics (2017)Google Scholar
  15. 15.
    Donahue, J., et al.: Long-term recurrent convolutional networks for visual recognition and description. IEEE Trans. Pattern Anal. Mach. Intell. 39(4), 677–691 (2017)CrossRefGoogle Scholar
  16. 16.
    EU-GDPR: Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation). Official J. Eur. Union L 119(59), 1–88 (2016)Google Scholar
  17. 17.
    Geiger, J.T., Zhang, Z., Weninger, F., Schuller, B., Rigoll, G.: Robust speech recognition using long short-term memory recurrent neural networks for hybrid acoustic modelling. In: Proceedings of the 15th Annual Conference of the International Speech Communication Association (INTERSPEECH), pp. 631–635 (2014)Google Scholar
  18. 18.
    Gers, F.A., Schmidhuber, J.: Recurrent nets that time and count. In: Proceedings of the IEEE International Joint Conference on Neural Networks (IJCNN), vol. 3, pp. 189–194 (2000)Google Scholar
  19. 19.
    Gers, F.A., Schmidhuber, J., Cummins, F.: Learning to forget: continual prediction with LSTM. In: Proceedings of the International Conference on Artificial Neural Networks (ICANN), vol. 2, pp. 850–855 (1999)Google Scholar
  20. 20.
    Gers, F.A., Schmidhuber, J., Cummins, F.: Learning to forget: continual prediction with LSTM. Neural Comput. 12(10), 2451–2471 (2000)CrossRefGoogle Scholar
  21. 21.
    Gevrey, M., Dimopoulos, I., Lek, S.: Review and comparison of methods to study the contribution of variables in artificial neural network models. Ecol. Model. 160(3), 249–264 (2003)CrossRefGoogle Scholar
  22. 22.
    Gonzalez-Dominguez, J., Lopez-Moreno, I., Sak, H., Gonzalez-Rodriguez, J., Moreno, P.J.: Automatic language identification using long short-term memory recurrent neural networks. In: Proceedings of the 15th Annual Conference of the International Speech Communication Association (INTERSPEECH), pp. 2155–2159 (2014)Google Scholar
  23. 23.
    Graves, A.: Generating sequences with recurrent neural networks. arXiv:1308.0850 (2014)
  24. 24.
    Graves, A., Liwicki, M., Fernandez, S., Bertolami, R., Bunke, H., Schmidhuber, J.: A novel connectionist system for unconstrained handwriting recognition. IEEE Trans. Pattern Anal. Mach. Intell. 31(5), 855–868 (2009)CrossRefGoogle Scholar
  25. 25.
    Graves, A., Schmidhuber, J.: Framewise phoneme classification with bidirectional LSTM and other neural network architectures. Neural Networks 18(5–6), 602–610 (2005)CrossRefGoogle Scholar
  26. 26.
    Greff, K., Srivastava, R.K., Koutník, J., Steunebrink, B.R., Schmidhuber, J.: LSTM: a search space odyssey. IEEE Trans. Neural Netw. Learn. Syst. 28(10), 2222–2232 (2017)MathSciNetCrossRefGoogle Scholar
  27. 27.
    Hausknecht, M., Stone, P.: Deep recurrent Q-learning for partially observable MDPs. In: AAAI Fall Symposium Series - Sequential Decision Making for Intelligent Agents, pp. 29–37 (2015)Google Scholar
  28. 28.
    Heess, N., Wayne, G., Tassa, Y., Lillicrap, T., Riedmiller, M., Silver, D.: Learning and transfer of modulated locomotor controllers. arXiv:1610.05182 (2016)
  29. 29.
    Hochreiter, S.: Implementierung und Anwendung eines ‘neuronalen’ Echtzeit-Lernalgorithmus für reaktive Umgebungen. Practical work, Institut für Informatik, Technische Universität München (1990)Google Scholar
  30. 30.
    Hochreiter, S.: Untersuchungen zu dynamischen neuronalen Netzen. Master’s thesis. Institut für Informatik, Technische Universität München (1991)Google Scholar
  31. 31.
    Hochreiter, S.: Recurrent neural net learning and vanishing gradient. In: Freksa, C. (ed.) Proceedings in Artificial Intelligence - Fuzzy-Neuro-Systeme 1997 Workshop, pp. 130–137. Infix (1997)Google Scholar
  32. 32.
    Hochreiter, S.: The vanishing gradient problem during learning recurrent neural nets and problem solutions. Int. J. Uncertainty Fuzziness Knowl. Based Syst. 6(2), 107–116 (1998)MathSciNetCrossRefGoogle Scholar
  33. 33.
    Hochreiter, S., Bengio, Y., Frasconi, P., Schmidhuber, J.: Gradient flow in recurrent nets: the difficulty of learning long-term dependencies. In: Kolen, J.F., Kremer, S.C. (eds.) A Field Guide to Dynamical Recurrent Networks, pp. 237–244. IEEE Press, New York (2001)Google Scholar
  34. 34.
    Hochreiter, S., Heusel, M., Obermayer, K.: Fast model-based protein homology detection without alignment. Bioinformatics 23(14), 1728–1736 (2007)CrossRefGoogle Scholar
  35. 35.
    Hochreiter, S., Schmidhuber, J.: Long short-term memory. Technical report, FKI-207-95, Fakultät für Informatik, Technische Universität München (1995)Google Scholar
  36. 36.
    Hochreiter, S., Schmidhuber, J.: LSTM can solve hard long time lag problems. In: Advances in Neural Information Processing Systems 9 (NIPS), pp. 473–479 (1996)Google Scholar
  37. 37.
    Hochreiter, S., Schmidhuber, J.: Long short-term memory. Neural Comput. 9(8), 1735–1780 (1997)CrossRefGoogle Scholar
  38. 38.
    Hochreiter, S., Younger, A.S., Conwell, P.R.: Learning to learn using gradient descent. In: Proceedings of the International Conference on Artificial Neural Networks (ICANN), pp. 87–94 (2001)CrossRefGoogle Scholar
  39. 39.
    Horst, F., Lapuschkin, S., Samek, W., Müller, K.R., Schöllhorn, W.I.: Explaining the unique nature of individual gait patterns with deep learning. Sci. Rep. 9, 2391 (2019)CrossRefGoogle Scholar
  40. 40.
    Kauffmann, J., Esders, M., Montavon, G., Samek, W., Müller, K.R.,: From clustering to cluster explanations via neural networks. arXiv:1906.07633 (2019)
  41. 41.
    Landecker, W., Thomure, M.D., Bettencourt, L.M.A., Mitchell, M., Kenyon, G.T., Brumby, S.P.: Interpreting individual classifications of hierarchical networks. In: IEEE Symposium on Computational Intelligence and Data Mining (CIDM), pp. 32–38 (2013)Google Scholar
  42. 42.
    Lapuschkin, S., Binder, A., Montavon, G., Müller, K.R., Samek, W.: The LRP toolbox for artificial neural networks. J. Mach. Learn. Res. 17(114), 1–5 (2016)MathSciNetzbMATHGoogle Scholar
  43. 43.
    Lapuschkin, S., Binder, A., Müller, K.R., Samek, W.: Understanding and comparing deep neural networks for age and gender classification. In: IEEE International Conference on Computer Vision Workshops, pp. 1629–1638 (2017)Google Scholar
  44. 44.
    Lapuschkin, S., Wäldchen, S., Binder, A., Montavon, G., Samek, W., Müller, K.R.: Unmasking clever hans predictors and assessing what machines really learn. Nat. Commun. 10, 1096 (2019)CrossRefGoogle Scholar
  45. 45.
    Li, J., Chen, X., Hovy, E., Jurafsky, D.: Visualizing and understanding neural models in NLP. In: Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT), pp. 681–691. Association for Computational Linguistics (2016)Google Scholar
  46. 46.
    Li, J., Monroe, W., Jurafsky, D.: Understanding neural networks through representation erasure. arXiv:1612.08220 (2017)
  47. 47.
    Lundberg, S.M., Lee, S.I.: A unified approach to interpreting model predictions. In: Advances in Neural Information Processing Systems 30 (NIPS), pp. 4765–4774 (2017)Google Scholar
  48. 48.
    Luoma, J., Ruutu, S., King, A.W., Tikkanen, H.: Time delays, competitive interdependence, and firm performance. Strateg. Manag. J. 38(3), 506–525 (2017)CrossRefGoogle Scholar
  49. 49.
    Marchi, E., Ferroni, G., Eyben, F., Gabrielli, L., Squartini, S., Schuller, B.: Multi-resolution linear prediction based features for audio onset detection with bidirectional LSTM neural networks. In: IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 2164–2168 (2014)Google Scholar
  50. 50.
    Mnih, V., et al.: Asynchronous methods for deep reinforcement learning. In: Proceedings of the 33rd International Conference on Machine Learning (ICML), vol. 48, pp. 1928–1937 (2016)Google Scholar
  51. 51.
    Montavon, G., Binder, A., Lapuschkin, S., Samek, W., Müller, K.-R.: Layer-wise relevance propagation: an overview. In: Samek, W. et al. (eds.) Explainable AI, LNCS 11700, pp. 193–209. Springer, Heidelberg (2019)Google Scholar
  52. 52.
    Montavon, G., Lapuschkin, S., Binder, A., Samek, W., Müller, K.R.: Explaining nonlinear classification decisions with deep Taylor decomposition. Pattern Recogn. 65, 211–222 (2017)CrossRefGoogle Scholar
  53. 53.
    Montavon, G., Samek, W., Müller, K.R.: Methods for interpreting and understanding deep neural networks. Digit. Signal Proc. 73, 1–15 (2018)MathSciNetCrossRefGoogle Scholar
  54. 54.
    Morcos, A.S., Barrett, D.G., Rabinowitz, N.C., Botvinick, M.: On the importance of single directions for generalization. In: International Conference on Learning Representations (ICLR) (2018)Google Scholar
  55. 55.
    Munro, P.: A dual back-propagation scheme for scalar reward learning. In: Proceedings of the Ninth Annual Conference of the Cognitive Science Society, pp. 165–176 (1987)Google Scholar
  56. 56.
    Murdoch, W.J., Liu, P.J., Yu, B.: Beyond word importance: contextual decomposition to extract interactions from LSTMs. In: International Conference on Learning Representations (ICLR) (2018)Google Scholar
  57. 57.
    Poerner, N., Schütze, H., Roth, B.: Evaluating neural network explanation methods using hybrid documents and morphosyntactic agreement. In: Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (ACL), pp. 340–350. Association for Computational Linguistics (2018)Google Scholar
  58. 58.
    Rahmandad, H., Repenning, N., Sterman, J.: Effects of feedback delay on learning. Syst. Dyn. Rev. 25(4), 309–338 (2009)CrossRefGoogle Scholar
  59. 59.
    Rieger, L., Chormai, P., Montavon, G., Hansen, L.K., Müller, K.-R.: Structuring neural networks for more explainable predictions. In: Escalante, H.J., et al. (eds.) Explainable and Interpretable Models in Computer Vision and Machine Learning. TSSCML, pp. 115–131. Springer, Cham (2018).  https://doi.org/10.1007/978-3-319-98131-4_5CrossRefGoogle Scholar
  60. 60.
    Robinson, A.J.: Dynamic error propagation networks. Ph.D. thesis, Trinity Hall and Cambridge University Engineering Department (1989)Google Scholar
  61. 61.
    Robinson, T., Fallside, F.: Dynamic reinforcement driven error propagation networks with application to game playing. In: Proceedings of the 11th Conference of the Cognitive Science Society, Ann Arbor, pp. 836–843 (1989)Google Scholar
  62. 62.
    Sahni, H.: Reinforcement learning never worked, and ‘deep’ only helped a bit. himanshusahni.github.io/2018/02/23/reinforcement-learning-never-worked.html (2018)
  63. 63.
    Sak, H., Senior, A., Beaufays, F.: Long short-term memory recurrent neural network architectures for large scale acoustic modeling. In: Proceedings of the 15th Annual Conference of the International Speech Communication Association (INTERSPEECH), Singapore, pp. 338–342 (2014)Google Scholar
  64. 64.
    Samek, W., Binder, A., Montavon, G., Lapuschkin, S., Müller, K.R.: Evaluating the visualization of what a deep neural network has learned. IEEE Trans. Neural Netw. Learn. Syst. 28(11), 2660–2673 (2017)MathSciNetCrossRefGoogle Scholar
  65. 65.
    Schmidhuber, J.: Making the world differentiable: on using fully recurrent self-supervised neural networks for dynamic reinforcement learning and planning in non-stationary environments. Technical report, FKI-126-90 (revised), Institut für Informatik, Technische Universität München (1990). Experiments by Sepp HochreiterGoogle Scholar
  66. 66.
    Schmidhuber, J.: Deep learning in neural networks: an overview. Neural Netw. 61, 85–117 (2015)CrossRefGoogle Scholar
  67. 67.
    Shrikumar, A., Greenside, P., Kundaje, A.: Learning important features through propagating activation differences. In: Proceedings of the 34th International Conference on Machine Learning (ICML), vol. 70, pp. 3145–3153 (2017)Google Scholar
  68. 68.
    Simonyan, K., Vedaldi, A., Zisserman, A.: Deep inside convolutional networks: visualising image classification models and saliency maps. In: International Conference on Learning Representations (ICLR) (2014)Google Scholar
  69. 69.
    Socher, R., et al.: Recursive deep models for semantic compositionality over a sentiment treebank. In: Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp. 1631–1642. Association for Computational Linguistics (2013)Google Scholar
  70. 70.
    Srivastava, N., Mansimov, E., Salakhudinov, R.: Unsupervised learning of video representations using LSTMs. In: Proceedings of the 32nd International Conference on Machine Learning (ICML), vol. 37, pp. 843–852 (2015)Google Scholar
  71. 71.
    Sturm, I., Lapuschkin, S., Samek, W., Müller, K.R.: Interpretable deep neural networks for single-trial EEG classification. J. Neurosci. Methods 274, 141–145 (2016)CrossRefGoogle Scholar
  72. 72.
    Sundararajan, M., Taly, A., Yan, Q.: Axiomatic attribution for deep networks. In: Proceedings of the 34th International Conference on Machine Learning (ICML), vol. 70, pp. 3319–3328 (2017)Google Scholar
  73. 73.
    Sutskever, I., Vinyals, O., Le, Q.V.: Sequence to sequence learning with neural networks. In: Advances in Neural Information Processing Systems 27 (NIPS), pp. 3104–3112 (2014)Google Scholar
  74. 74.
    Sutton, R.S., Barto, A.G.: Reinforcement Learning: An Introduction, 2nd edn. MIT Press, Cambridge (2017). Draft from November 2017zbMATHGoogle Scholar
  75. 75.
    Thuillier, E., Gamper, H., Tashev, I.J.: Spatial audio feature discovery with convolutional neural networks. In: IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 6797–6801 (2018)Google Scholar
  76. 76.
    Venugopalan, S., Xu, H., Donahue, J., Rohrbach, M., Mooney, R., Saenko, K.: Translating videos to natural language using deep recurrent neural networks. In: Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT), pp. 1494–1504. Association for Computational Linguistics (2015)Google Scholar
  77. 77.
    Yang, Y., Tresp, V., Wunderle, M., Fasching, P.A.: Explaining therapy predictions with layer-wise relevance propagation in neural networks. In: IEEE International Conference on Healthcare Informatics (ICHI), pp. 152–162 (2018)Google Scholar
  78. 78.
    Zaremba, W., Sutskever, I., Vinyals, O.: Recurrent neural network regularization. arXiv:1409.2329 (2015)
  79. 79.
    Zeiler, M.D., Fergus, R.: Visualizing and understanding convolutional networks. In: Fleet, D., Pajdla, T., Schiele, B., Tuytelaars, T. (eds.) ECCV 2014. LNCS, vol. 8689, pp. 818–833. Springer, Cham (2014).  https://doi.org/10.1007/978-3-319-10590-1_53CrossRefGoogle Scholar
  80. 80.
    Zhang, J., Lin, Z., Brandt, J., Shen, X., Sclaroff, S.: Top-down neural attention by excitation backprop. In: Leibe, B., Matas, J., Sebe, N., Welling, M. (eds.) ECCV 2016. LNCS, vol. 9908, pp. 543–559. Springer, Cham (2016).  https://doi.org/10.1007/978-3-319-46493-0_33CrossRefGoogle Scholar

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  • Leila Arras
    • 1
  • José Arjona-Medina
    • 2
  • Michael Widrich
    • 2
  • Grégoire Montavon
    • 3
  • Michael Gillhofer
    • 2
  • Klaus-Robert Müller
    • 3
    • 4
    • 5
  • Sepp Hochreiter
    • 2
  • Wojciech Samek
    • 1
    Email author
  1. 1.Fraunhofer Heinrich Hertz InstituteBerlinGermany
  2. 2.Johannes Kepler University LinzLinzAustria
  3. 3.Technische Universität BerlinBerlinGermany
  4. 4.Korea UniversitySeongbuk-gu, SeoulKorea
  5. 5.Max Planck Institute for InformaticsSaarbrückenGermany

Personalised recommendations