26,567 Hits in 4.3 sec

Explaining Convolutional Neural Networks using Softmax Gradient Layer-wise Relevance Propagation [article]

Brian Kenji Iwana, Ryohei Kuroki, Seiichi Uchida
2019 arXiv   pre-print
Specifically, we propose a novel visualization method of pixel-wise input attribution called Softmax-Gradient Layer-wise Relevance Propagation (SGLRP).  ...  We confirm that SGLRP performs better than existing Layer-wise Relevance Propagation (LRP) based methods and can help in the understanding of the decision process of CNNs.  ...  Relevance Propagation Layer-wise Relevance Propagation LRP [3] is based on the idea that the likelihood of a class can be traced backwards through a network to the individual layer-wise nodes or elements  ... 
arXiv:1908.04351v3 fatcat:lg7zmadpwbfmhodaa2blq735rm

SLRP: Improved heatmap generation via selective layer‐wise relevance propagation

Yeon‐Jee Jung, Seung‐Ho Han, Ho‐Jin Choi
2021 Electronics Letters  
However, even advanced versions of layer-wise relevance propagation (such as contrastive layer-wise relevance propagation and softmax-gradient layer-wise relevance propagation) have some limitations.  ...  A typical approach is layer-wise relevance propagation, which generates a heatmap, where each pixel value represents the contributions to the model's predictions.  ...  This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.  ... 
doi:10.1049/ell2.12061 fatcat:pcjqqd6jpng3zkhfytzisg46zi

Beyond saliency: understanding convolutional neural networks from saliency prediction on layer-wise relevance propagation

Heyi Li, Yunke Tian, Klaus Mueller, Xin Chen
2019 Image and Vision Computing  
Our proposed method starts out with a layer-wise relevance propagation (LRP) step which estimates a pixel-wise relevance map over the input image.  ...  As such, our proposed SR map constitutes a convenient visual interface which unveils the visual attention of the network and reveals which type of objects the model has learned to recognize after training  ...  Tengyu Ma for his comments on an earlier version of the manuscript.  ... 
doi:10.1016/j.imavis.2019.02.005 fatcat:4gumh6ftkjgkxfr7q3ktg63epq

Feature visualization within an automated design assessment leveraging explainable artificial intelligence methods

Raoul Schönhof, Artem Werner, Jannes Elstner, Boldizsar Zopcsak, Ramez Awad, Marco Huber
2021 Procedia CIRP  
Within this work, a sensitivity analysis (SA), the layer-wise relevance propagation (LRP), the Gradient-weighted Class Activation Mapping (Grad-CAM) method as well as the Local Interpretable Model-Agnostic  ...  In the medium run, this might enable to identify regions of interest supporting product designers to optimize their models with regards to assembly processes.  ...  Layer-Wise Relevance Propagation 3D When handling 3D models, Layer-wise Relevance Propagation comes very handy.  ... 
doi:10.1016/j.procir.2021.05.075 fatcat:cipszsstfveftlxfcieztvitcy

Visualizing and Understanding Neural Machine Translation

Yanzhuo Ding, Yang Liu, Huanbo Luan, Maosong Sun
2017 Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)  
In this work, we propose to use layer-wise relevance propagation (LRP) to compute the contribution of each contextual word to arbitrary hidden states in the attention-based encoderdecoder framework.  ...  We show that visualization with LRP helps to interpret the internal workings of NMT and analyze translation errors.  ...  Acknowledgements This work is supported by the National Natural Science Foundation of China (No.61522204), the 863 Program (2015AA015407), and the National Natural Science Foundation of China (No.61432013  ... 
doi:10.18653/v1/p17-1106 dblp:conf/acl/DingLLS17 fatcat:doukf6zz25hkhmhrkato5etrle

On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation

Sebastian Bach, Alexander Binder, Grégoire Montavon, Frederick Klauschen, Klaus-Robert Müller, Wojciech Samek, Oscar Deniz Suarez
2015 PLoS ONE  
This work proposes a general solution to the problem of understanding classification decisions by pixel-wise decomposition of nonlinear classifiers.  ...  We introduce a methodology that allows to visualize the contributions of single pixels to predictions for kernel-based classifiers over Bag of Words features and for multilayered neural networks.  ...  KRM thanks for partial funding by the National Research Foundation of Korea funded by the Ministry of Education, Science, and Technology in the BK21 program.  ... 
doi:10.1371/journal.pone.0130140 pmid:26161953 pmcid:PMC4498753 fatcat:gfb4q6cdirggbd7auywpploz6m

How do Convolutional Neural Networks Learn Design? [article]

Shailza Jolly, Brian Kenji Iwana, Ryohei Kuroki, Seiichi Uchida
2018 arXiv   pre-print
In order to understand these visual clues contributing towards the decision of a genre, we present the application of Layer-wise Relevance Propagation (LRP) on the book cover image classification results  ...  We use LRP to explain the pixel-wise contributions of book cover design and highlight the design elements contributing towards particular genres.  ...  To interpret the reasoning behind a CNN's prediction we used a method called Layer-wise Relevance Propagation (LRP) [7] .  ... 
arXiv:1808.08402v1 fatcat:yetuivru7bat3gcawk72uinupy

Visualizing and Understanding Patch Interactions in Vision Transformer [article]

Jie Ma, Yalong Bai, Bineng Zhong, Wei Zhang, Ting Yao, Tao Mei
2022 arXiv   pre-print
Specifically, we first introduce a quantification indicator to measure the impact of patch interaction and verify such quantification on attention window design and indiscriminative patches removal.  ...  Extensive experiments on ImageNet demonstrate that the exquisitely designed quantitative method is shown able to facilitate ViT model learning, leading the top-1 accuracy by 4.28% at most.  ...  [36] apply layer-wise relevance propagation to consider the different relevance of multihead attention block.  ... 
arXiv:2203.05922v1 fatcat:sc5cmsniqrdvjnhyan7fzin6iy

Identification and Visualization of the Underlying Independent Causes of the Diagnostic of Diabetic Retinopathy made by a Deep Learning Classifier [article]

Jordi de la Torre, Aida Valls, Domenec Puig, Pere Romero-Aroca
2018 arXiv   pre-print
Interpretability is a key factor in the design of automatic classifiers for medical diagnosis.  ...  We use a combination of Independent Component Analysis with a Score Visualization technique.  ...  A posteriori we use a pixel-wise relevance propagation derived method to visualize such independent component in input space.  ... 
arXiv:1809.08567v1 fatcat:kl56vevr3vbkxd5odmr2hrhnba

Interpretable Feature Learning Framework for Smoking Behavior Detection [article]

Nakayiza Hellen, Ggaliwango Marvin
2021 arXiv   pre-print
Relevance Propagation (LRP) to explain the network detection or prediction of smoking behavior based on the most relevant learned features or pixels or neurons.  ...  We developed an Interpretable feature learning framework for smoking behavior detection which utilizes a Deep Learning VGG-16 pretrained network to predict and classify the input Image class and a Layer-wise  ...  Layer-Wise Relevance Propagation (LRP). Bach et al.  ... 
arXiv:2112.08178v1 fatcat:x62ul3kdv5dijb3gtqoz3f6mnu

Explaining nonlinear classification decisions with deep Taylor decomposition

Grégoire Montavon, Sebastian Lapuschkin, Alexander Binder, Wojciech Samek, Klaus-Robert Müller
2017 Pattern Recognition  
Although these methods perform impressively well, they have a significant disadvantage, the lack of transparency, limiting the interpretability of the solution and thus the scope of application in practice  ...  Our method is based on deep Taylor decomposition and efficiently utilizes the structure of the network by backpropagating the explanations from the output to the input layer.  ...  Fig. 2 illustrates the procedure of layer-wise relevance propagation on a cartoon example where an image of a cat is presented to a deep network.  ... 
doi:10.1016/j.patcog.2016.11.008 fatcat:mf3ycvzqvbfupp5oy53apleyku

A Deep Learning Interpretable Classifier for Diabetic Retinopathy Disease Grading

Jordi de la Torre, Aida Valls, Domenec Puig
2019 Neurocomputing  
The vast amount of parameters of these models make difficult to infer a rationale interpretation from them.  ...  The generated visual maps can be interpreted by an expert in order to compare its own knowledge with the interpretation given by the model.  ...  The model is based on the layer-wise relevance propagation model described above. We reformulate one of the properties of the relevance propagation.  ... 
doi:10.1016/j.neucom.2018.07.102 fatcat:h4yrltm4tfg3nhidua33saab5e

Interpreting Galaxy Deblender GAN from the Discriminator's Perspective [article]

Heyi Li, Yuewei Lin, Klaus Mueller, Wei Xu
2020 arXiv   pre-print
This research focuses on behaviors of one of the network's major components, the Discriminator, which plays a vital role but is often overlooked, Specifically, we enhance the Layer-wise Relevance Propagation  ...  We find that our proposed method serves as a useful visual analytical tool for a deeper understanding of GAN models.  ...  Among the limited works explaining generative network models, Liu [16] designed a GUI interface to display connections between neurons of neighboring layers in a GAN model.  ... 
arXiv:2001.06151v1 fatcat:si4goztr5fdclip7v3i2ulrulu

Towards Demystifying Subliminal Persuasiveness: Using XAI-Techniques to Highlight Persuasive Markers of Public Speeches [chapter]

Klaus Weber, Lukas Tinnes, Tobias Huber, Alexander Heimerl, Marc-Leon Reinecker, Eva Pohlen, Elisabeth André
2020 Lecture Notes in Computer Science  
We then created visualizations of the predictions by making use of the explainable artificial intelligence methods Grad-CAM and layer-wise relevance propagation that highlight the most relevant image sections  ...  and trained a neural network capable of predicting the degree of perceived convincingness based on visual input only.  ...  work has been funded by the Deutsche Forschungsgemeinschaft (DFG) within the project "How to Win Arguments -Empowering Virtual Agents to Improve their Persuasiveness", Grant Number 376696351, as part of  ... 
doi:10.1007/978-3-030-51924-7_7 fatcat:2dmmt6dpkvgb3njsf2rxry3fya

Controlling Explanatory Heatmap Resolution and Semantics via Decomposition Depth [article]

Sebastian Bach, Alexander Binder, Klaus-Robert Müller, Wojciech Samek
2016 arXiv   pre-print
Layer-wise Relevance Propagation (LRP) is a method to compute scores for individual components of an input image, denoting their contribution to the prediction of the classifier for one particular test  ...  We present an application of the Layer-wise Relevance Propagation (LRP) algorithm to state of the art deep convolutional neural networks and Fisher Vector classifiers to compare the image perception and  ...  LAYER-WISE RELEVANCE PROPAGATION The aim of LRP is to attribute shares of upper layer relevances R (l+1) j to all components i of the adjacent lower layer l, such that each component of l receives a relevance  ... 
arXiv:1603.06463v3 fatcat:taoyrwrtnbf55f2bgc7iafqole
« Previous Showing results 1 — 15 out of 26,567 results