Grégoire Montavon

Grégoire Montavon is a senior researcher in the Machine Learning Group at the Technische Universität Berlin, and in the Berlin Institute for the Foundations of Learning and Data (BIFOLD). He received a Masters degree in Communication Systems from École Polytechnique Fédérale de Lausanne in 2009, and a Ph.D. degree in Machine Learning from the Technische Universität Berlin in 2013. His research interests include explainable AI, deep neural networks, and unsupervised learning.

The goal of this project is to develop methods to explain the decisions (e.g. classifications) of machine learning models in terms of input variables (i.e. which input variables are relevant for the model's decision).

• Explanations in deep neural networks is particularly difficult due to the high level of nonlinearity of the decision function. We have proposed a procedure called Layer-Wise Relevance Propagation (LRP) [4] that robustly explains deep neural networks at low computational cost.
• We have proposed the Deep Taylor Decomposition [5] framework, which connects mathematically the LRP procedure to Taylor expansions. Specifically, the LRP algorithm can be interpreted as a collection of Taylor expansions performed locally in the network, and this connection allows us to systematically bring LRP to new architectures with different layer types.
• In order to apply explanation techniques beyond neural networks, we have proposed a Neuralization-Propagation [6] approach which consists of (1) rewriting non neural network models such as kernel-based anomaly detection or k-means clustering as strictly equivalent neural networks, and (2) use the neural network representation and LRP to produce the explanation. Unlike surrogate modeling approaches, our neuralization-propagation approach does not require any retraining.
• Because a simple attribution of the prediction on the input features may not be sufficient for certain tasks (e.g. predicting similarity or predicting graphs), we have developed extensions for these cases: BiLRP [7] and GNN-LRP [8], which produce explanations in terms of pairs of input features or in terms of walks into the graph.
• The methods we have developed can serve to verify that trained neural networks predict as expected (i.e. are not subject to a Clever Hans [9] effect), and we are also starting to apply our method to address challenges in the area of digital humanities and quantum chemistry.

Demos, tutorials, software, and list of publications are available at www.heatmapping.org [10]

Winter semester 2020/2021:

• Machine Learning 1 [11]

Summer semester 2020:

• Machine Learning 2 [12]

Preprints

• L Ruff, J Kauffmann, R Vandermeulen, G Montavon, W Samek, M Kloft, T Dietterich, KR Müller
  A Unifying Review of Deep and Shallow Anomaly Detection [13]
  CoRR abs/2009.11732 (2020)
• J Kauffmann, L Ruff, G Montavon, KR Müller
  The Clever Hans Effect in Anomaly Detection [14]
  CoRR abs/2006.10609 (2020)
• T Schnake, O Eberle, J Lederer, S Nakajima, K T. Schütt, KR Müller, G Montavon
  Higher-Order Explanations of Graph Neural Networks via Relevant Walks [15]
  CoRR abs/2006.03589 (2020)
• W Samek, G Montavon, S Lapuschkin, C Anders, KR Müller
  Toward Interpretable Machine Learning: Transparent Deep Neural Networks and Beyond [16]
  CoRR abs/2003.07631 (2020)
• J Kauffmann, M Esders, G Montavon, W Samek, KR Müller
  From Clustering to Cluster Explanations via Neural Networks [17]
  CoRR abs/1906.07633 (2019)

 Edited books

• W Samek, G Montavon, A Vedaldi, LK Hansen, KR Müller (Eds.)
  Explainable AI: Interpreting, Explaining and Visualizing Deep Learning [18]
  Springer LNCS 11700 (2019)
• G Montavon, G Orr, KR Müller (Eds.)
  Neural Networks: Tricks of the Trade, 2nd Edn [19]
  Springer LNCS 7700 (2012)

 Book chapters

• G Montavon
  Introduction to Neural Networks [20]
  Machine Learning Meets Quantum Physics, 37-62 (2020)
• G Montavon, A Binder, S Lapuschkin, W Samek, KR Müller
  Layer-Wise Relevance Propagation: An Overview [21]
  in Explainable AI, Springer LNCS 11700 (2019)
• L Arras, J Arjona, M Widrich, G Montavon, M Gillhofer, KR Müller, S Hochreiter, W Samek
  Explaining and Interpreting LSTMs [22]
  in Explainable AI, Springer LNCS 11700 (2019)
• G Montavon
  Gradient-Based Vs. Propagation-Based Explanations: An Axiomatic Comparison [23]
  in Explainable AI, Springer LNCS 11700 (2019)
• C Anders, G Montavon, W Samek, KR Müller
  Understanding Patch-Based Learning of Video Data by Explaining Predictions [24]
  in Explainable AI, Springer LNCS 11700 (2019)
• L Rieger, P Chormai, G Montavon, LK Hansen, KR Müller
  Structuring Neural Networks for More Explainable Predictions [25]
  in Explainable and Interpretable Models in Computer Vision and Machine Learning, pp 115-131, Springer SSCML (2018)
• G Montavon, KR Müller
  Deep Boltzmann Machines and the Centering Trick [26]
  in Neural Networks: Tricks of the Trade, 2nd Edn, pp 621-637, Springer LNCS, vol. 7700 (2012)

Journal Publications

• K Melnyk, S Klus, G Montavon, T Conrad
  GraphKKE: Graph Kernel Koopman Embedding for Human Microbiome Analysis [27]
  Applied Network Science (accepted)
• O Eberle, J Büttner, F Kräutli, KR Müller, M Valleriani, G Montavon
  Building and Interpreting Deep Similarity Models [28]
  IEEE Transactions on Pattern Analysis and Machine Intelligence (2020)
• S Agarwal, N Tosi, D Breuer, S Padovan, P Kessel, G Montavon
  A machine-learning-based surrogate model of Mars’ thermal evolution [29]
  Geophysical Journal International, ggaa234 (2020)
• P Baumeister, S Padovan, N Tosi, G Montavon, N Nettelmann, J MacKenzie, M Godolt
  Machine learning inference of the interior structure of low-mass exoplanets [30]
  The Astrophysical Journal, 118(1):42 (2020)
• J Kauffmann, KR Müller, G Montavon
  Towards Explaining Anomalies: A Deep Taylor Decomposition of One-Class Models [31]
  Pattern Recognition, 107198 (2020)
• P Jurmeister, M Bockmayr, P Seegerer, T Bockmayr,D Treue, G Montavon, C Vollbrecht, A Arnold, D Teichmann, K Bressem, U Schüller, Mv Laffert, KR Müller, D Capper, F Klauschen
  Machine learning analysis of DNA methylation profiles distinguishes primary lung squamous cell carcinomas from head and neck metastases [32]
  Science Translational Medicine 11(509):eaaw8513 (2019)
• S Lapuschkin, S Wäldchen, A Binder, G Montavon, W Samek, KR Müller
  Unmasking Clever Hans Predictors and Assessing What Machines Really Learn [33]
  Nature Communications 10:1096 (2019)
• D Heim, G Montavon, P Hufnagl, KR Müller, F Klauschen
  Computational analysis reveals histotype-dependent molecular profile and actionable mutation effects across cancers [34]
  Genome Medicine, 10:83 (2018)
• G Montavon, W Samek, KR Müller
  Methods for Interpreting and Understanding Deep Neural Networks [35]
  Digital Signal Processing, 73:1-15 (2018)
• L Arras, F Horn, G Montavon, KR Müller, W Samek
  "What is Relevant in a Text Document?": An Interpretable Machine Learning Approach [36]
  PLOS ONE, 12(8):e0181142 (2017)
• G Montavon, S Lapuschkin, A Binder, W Samek, KR Müller
  Explaining NonLinear Classification Decisions with Deep Taylor Decomposition [37]
  Pattern Recognition, 65:211–222 (2017)
• W Samek, A Binder, G Montavon, S Lapuschkin, KR Müller
  Evaluating the Visualization of What a Deep Neural Network has Learned [38]
  IEEE Transactions on Neural Networks and Learning Systems, 28(11):2660-2673 (2016)
• S Bach, A Binder, G Montavon, F Klauschen, KR Müller, W Samek
  On Pixel-wise Explanations for Non-Linear Classifier Decisions by Layer-wise Relevance Propagation [39]
  PLOS ONE, 10(7):e0130140 (2015)
• G Montavon, M Rupp, V Gobre, A Vazquez-Mayagoitia, K Hansen, A Tkatchenko, KR Müller, OAv Lilienfeld
  Machine Learning of Molecular Electronic Properties in Chemical Compound Space [40]
  New Journal of Physics, 15 095003 (2013)
• K Hansen, G Montavon, F Biegler, S Fazli, M Rupp, M Scheffler, OAv Lilienfeld, A Tkatchenko, KR Müller
  Assessment and Validation of Machine Learning Methods for Predicting Molecular Atomization Energies [41]
  Journal of Chemical Theory and Computation, 9(8):3404-3419 (2013)
• G Montavon, M Braun, T Krueger, KR Müller
  Analyzing Local Structure in Kernel-based Learning: Explanation, Complexity and Reliability Assessment [42]
  IEEE Signal Processing Magazine, 30(4):62-74 (2013)
• G Montavon, M Braun, KR Müller
  Kernel Analysis of Deep Networks [43]
  Journal of Machine Learning Research, 12:2563-2581 (2011)

Conference Publications

• G Montavon, KR Müller, M Cuturi
  Wasserstein Training of Restricted Boltzmann Machines [44]
  NIPS 2016
• F Arbabzadah, G Montavon, KR Müller, W Samek
  Identifying Individual Facial Expressions by Deconstructing a Neural Network [45]
  GCPR 2016
• A Binder, G Montavon, S Bach, KR Müller, W Samek
  Layer-wise Relevance Propagation for Neural Networks with Local Renormalization Layers [46]
  ICANN 2016
• S Lapuschkin, A Binder, G Montavon, KR Müller, W Samek
  Analyzing Classifiers: Fisher Vectors and Deep Neural Networks [47]
  CVPR 2016
• A Binder, S Bach, G Montavon, KR Müller, W Samek
  Layer-wise Relevance Propagation for Deep Neural Network Architectures [48]
  ICISA 2016
• G Montavon, K Hansen, S Fazli, M Rupp, F Biegler, A Ziehe, A Tkatchenko, OAv Lilienfeld, KR Müller
  Learning Invariant Representations of Molecules for Atomization Energy Prediction [49]
  NIPS 2012
• G Montavon, M Braun, KR Müller
  Deep Boltzmann Machines as Feed-Forward Hierarchies [50]
  AISTATS 2012
• G Montavon, M Braun, KR Müller
  Layer-Wise Analysis of Deep Networks with Gaussian Kernels [51]
  NIPS 2010

Software Publications

• M Alber, S Lapuschkin, P Seegerer, M Hägele, KT Schütt, G Montavon, W Samek, KR Müller, S Dähne, PJ Kindermans
  iNNvestigate neural networks! [52]
  Journal of Machine Learning Research, Software Track (2019)
• S Lapuschkin, A Binder, G Montavon, KR Müller, W Samek
  The Layer-wise Relevance Propagation Toolbox for Artificial Neural Networks [53]
  Journal of Machine Learning Research, Software Track (2016)

Workshop Publications

• L Arras, G Montavon, KR Müller, W Samek
  Explaining Recurrent Neural Network Predictions in Sentiment Analysis [54]
  EMNLP Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis (2017)
• W Samek, G Montavon, A Binder, S Lapuschkin, KR Müller
  Interpreting the Predictions of Complex ML Models by Layer-wise Relevance Propagation [55]
  NIPS Workshop on Interpretable ML for Complex Systems (2016)
• L Arras, F Horn, G Montavon, KR Müller, W Samek
  Explaining Predictions of Non-Linear Classifiers in NLP [56]
  ACL Workshop on Representation Learning for NLP (2016)
• G Montavon, S Bach, A Binder, W Samek, KR Müller
  Deep Taylor Decomposition of Neural Networks [57]
  ICML Workshop on Visualization for Deep Learning (2016)
• A Binder, W Samek, G Montavon, S Bach, KR Müller
  Analyzing and Validating Neural Networks Predictions [58]
  ICML Workshop on Visualization for Deep Learning (2016)
• G Montavon, KR Müller
  Neural Networks for Computational Chemistry: Pitfalls and Recommendations [59]
  MRS Online Proceedings Library (2013)
• G Montavon
  Deep Learning for Spoken Language Identification [60]
  NIPS Workshop on Deep Learning for Speech Recognition and Related Applications (2009)

Theses

• G Montavon
  On Layer-Wise Representations in Deep Neural Networks [61]
  PhD Thesis, Technische Universität Berlin, Germany (2013)
• G Montavon
  A Machine Learning Approach to Classification of Low Resolution Histological Samples [62]
  Master Thesis, École Polytechnique Fédérale de Lausanne, Switzerland (2009)

Unpublished

• J Kauffmann, G Montavon, LA Lima, S Nakajima, KR Müller, N Görnitz
  Unsupervised Detection and Explanation of Latent-class Contextual Anomalies [63] (2018)
• F Horn, L Arras, G Montavon, KR Müller, W Samek
  Exploring text datasets by visualizing relevant words [64] (2017)