TY - GEN
T1 - Layer-wise relevance propagation for neural networks with local renormalization layers
AU - Binder, Alexander
AU - Montavon, Grégoire
AU - Lapuschkin, Sebastian
AU - Müller, Klaus Robert
AU - Samek, Wojciech
N1 - Funding Information:
Part of the research was supported by the grant from the EC, HYPHEN Contract Nr: FP6-027455.
Publisher Copyright:
© Springer International Publishing Switzerland 2016.
PY - 2016
Y1 - 2016
N2 - Layer-wise relevance propagation is a framework which allows to decompose the prediction of a deep neural network computed over a sample, e.g. an image, down to relevance scores for the single input dimensions of the sample such as subpixels of an image. While this approach can be applied directly to generalized linear mappings, product type non-linearities are not covered. This paper proposes an approach to extend layer-wise relevance propagation to neural networks with local renormalization layers, which is a very common product-type non-linearity in convolutional neural networks. We evaluate the proposed method for local renormalization layers on the CIFAR-10, Imagenet and MIT Places datasets.
AB - Layer-wise relevance propagation is a framework which allows to decompose the prediction of a deep neural network computed over a sample, e.g. an image, down to relevance scores for the single input dimensions of the sample such as subpixels of an image. While this approach can be applied directly to generalized linear mappings, product type non-linearities are not covered. This paper proposes an approach to extend layer-wise relevance propagation to neural networks with local renormalization layers, which is a very common product-type non-linearity in convolutional neural networks. We evaluate the proposed method for local renormalization layers on the CIFAR-10, Imagenet and MIT Places datasets.
KW - Image classification
KW - Interpretability
KW - Neural networks
UR - http://www.scopus.com/inward/record.url?scp=84988311277&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-44781-0_8
DO - 10.1007/978-3-319-44781-0_8
M3 - Conference contribution
AN - SCOPUS:84988311277
SN - 9783319447803
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 63
EP - 71
BT - Artificial Neural Networks and Machine Learning - 25th International Conference on Artificial Neural Networks, ICANN 2016, Proceedings
A2 - Villa, Alessandro E.P.
A2 - Masulli, Paolo
A2 - Rivero, Antonio Javier Pons
PB - Springer Verlag
T2 - 25th International Conference on Artificial Neural Networks and Machine Learning, ICANN 2016
Y2 - 6 September 2016 through 9 September 2016
ER -