Layer-Wise Relevance Propagation: An Overview

Montavon, Grégoire; Binder, Alexander; Lapuschkin, Sebastian; Samek, Wojciech; Müller, Klaus Robert

doi:10.1007/978-3-030-28954-6_10

Cited by 523 publications

(472 citation statements)

References 38 publications

Supporting

Mentioning

468

Contrasting

Unclassified

Order By: Relevance

“…Such explanations help to verify the predictions and establish trust in the correct functioning on the system. Layer-wise Relevance Propagation (LRP) [9,58] provides a general framework for explaining individual predictions, i.e., it is applicable to various ML models, including neural networks [9], LSTMs [7], Fisher Vector classifiers [44] and Support Vector Machines [35]. Section 4 gives an overview over recently proposed methods for computing individual explanations.…”

Section: Explaining Individual Predictionsmentioning

confidence: 99%

Towards Explainable Artificial Intelligence

Samek

Müller

2019

Lecture Notes in Computer Science

Self Cite

559

357

View full text Add to dashboard Cite

In recent years, machine learning (ML) has become a key enabling technology for the sciences and industry. Especially through improvements in methodology, the availability of large databases and increased computational power, today's ML algorithms are able to achieve excellent performance (at times even exceeding the human level) on an increasing number of complex tasks. Deep learning models are at the forefront of this development. However, due to their nested non-linear structure, these powerful models have been generally considered "black boxes", not providing any information about what exactly makes them arrive at their predictions. Since in many applications, e.g., in the medical domain, such lack of transparency may be not acceptable, the development of methods for visualizing, explaining and interpreting deep learning models has recently attracted increasing attention. This introductory paper presents recent developments and applications in this field and makes a plea for a wider use of explainable learning algorithms in practice.

show abstract

Section: Explaining Individual Predictionsmentioning

confidence: 99%

Towards Explainable Artificial Intelligence

Samek

Müller

2019

Lecture Notes in Computer Science

Self Cite

559

357

View full text Add to dashboard Cite

show abstract

“…(a) Due to its attention-based design, a trained model can be used to compute the attention weights of a sequence, which directly indicates its importance. (b) DeepRC furthermore allows for the usage of contribution analysis methods, such as Integrated Gradients (IG) (Sundararajan et al, 2017) or Layer-Wise Relevance Propagation (Montavon et al, 2018;Arras et al, 2019;Montavon et al, 2019;Preuer et al, 2019). We apply IG to identify the input patterns that are relevant for the classification.…”

Section: A8 Interpreting Deeprcmentioning

confidence: 99%

Modern Hopfield Networks and Attention for Immune Repertoire Classification

Widrich

Schäfl

Pavlović

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…LRP has been most commonly applied to deep rectifier networks. In these networks, the activations at the current layer can be computed from activations in the previous layer as: a k = max 0, 0,j a j w jk A general family of propagation rules for such types of layer is given by [51]:…”

Section: Lrp In Deep Neuralmentioning

confidence: 99%

“…Specific propagation rules such as LRP-, LRP-α 1 β 0 and LRP-γ fall under this umbrella. They are easy to implement [42,51] and can be interpreted as the result of a deep Taylor decomposition of the neural network function [52]. On convolutional neural networks for computer vision, composite strategies making use of different rules at different layers have shown to work well in practice [43,51].…”

Section: Lrp In Deep Neuralmentioning

confidence: 99%

“…They are easy to implement [42,51] and can be interpreted as the result of a deep Taylor decomposition of the neural network function [52]. On convolutional neural networks for computer vision, composite strategies making use of different rules at different layers have shown to work well in practice [43,51]. An alternative default strategy in computer vision is to uniformly employ the LRP-α 1 β 0 in every hidden layer [53], the latter has the advantage of having no free parameter, and delivers positive explanations.…”

Section: Lrp In Deep Neuralmentioning

confidence: 99%

See 1 more Smart Citation

Explaining and Interpreting LSTMs

Arras

Arjona-Medina

Widrich

et al. 2019

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

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 L. Arras and J. Arjona-Medina contributed equally to this work.The final authenticated publication is available online at https://doi.

show abstract

Layer-Wise Relevance Propagation: An Overview

Cited by 523 publications

References 38 publications

Towards Explainable Artificial Intelligence

Towards Explainable Artificial Intelligence

Modern Hopfield Networks and Attention for Immune Repertoire Classification

Explaining and Interpreting LSTMs

Contact Info

Product

Resources

About