paper-with-me

홈 › Papers

Exemplary Natural Images Explain CNN Activations Better than State-of-the-Art Feature Visualization

2020-10-23 · Judy Borowski, Roland S. Zimmermann, Judith Schepers, Robert Geirhos, Thomas S. A. Wallis, Matthias Bethge, Wieland Brendel

Feature visualizations such as synthetic maximally activating images are a widely used explanation method to better understand the information processing of convolutional neural networks (CNNs). At the same time, there are concerns that these visualizations might not accurately represent CNNs' inner workings. Here, we measure how much extremely activating images help humans to predict CNN activations. Using a well-controlled psychophysical paradigm, we compare the informativeness of synthetic images by Olah et al. (2017) with a simple baseline visualization, namely exemplary natural images that also strongly activate a specific feature map. Given either synthetic or natural reference images, human participants choose which of two query images leads to strong positive activation. The experiments are designed to maximize participants' performance, and are the first to probe intermediate instead of final layer representations. We find that synthetic images indeed provide helpful information about feature map activations ($82\pm4\%$ accuracy; chance would be $50\%$). However, natural images - originally intended as a baseline - outperform synthetic images by a wide margin ($92\pm2\%$). Additionally, participants are faster and more confident for natural images, whereas subjective impressions about the interpretability of the feature visualizations are mixed. The higher informativeness of natural images holds across most layers, for both expert and lay participants as well as for hand- and randomly-picked feature visualizations. Even if only a single reference image is given, synthetic images provide less information than natural images ($65\pm5\%$ vs. $73\pm4\%$). In summary, synthetic images from a popular feature visualization method are significantly less informative for assessing CNN activations than natural images. We argue that visualization methods should improve over this baseline.

📄 PDF Abstract BibTeX arXiv:2010.12606

Code (1)

bethgelab/testing_visualizations 공식 구현 tf

Tasks

Informativeness

Methods 이 논문이 사용한 방법론

Interpretability 설명 없음

Similar Papers 제목 키워드 기반

Exemplary natural images explain CNN activations better than synthetic feature visualizations

2021-01-01 · ICLR 2021 1 · Judy Borowski, Roland Simon Zimmermann, Judith Schepers, Robert Geirhos 외

Feature visualizations such as synthetic maximally activating images are a widely used explanation method to better understand the information processing of convolutional neural networks (CNNs). At the same time, there a…

Informativeness

Explainable AI for Engineering Design: A Unified Approach of Systems Engineering and Component- Based Deep Learning Demonstrated by Energy- Efficient Building Design

2021-08-30 · Philipp Geyer, Manav Mahan Singh, Xia Chen

Data-driven models created by machine learning, gain in importance in all fields of design and engineering. They, have high potential to assist decision-makers in creating novel, artefacts with better performance and sus…

Natural Images are More Informative for Interpreting CNN Activations than State-of-the-Art Synthetic Feature Visualizations

2020-10-09 · NeurIPS Workshop SVRHM 2020 12 · Judy Borowski, Roland Simon Zimmermann, Judith Schepers, Robert Geirhos 외

Feature visualizations such as synthetic maximally activating images are a widely used explanation method to better understand the information processing of convo- lutional neural networks (CNNs). At the same time, there…

Informativeness

Conditioning Sequence-to-sequence Networks with Learned Activations

2021-09-29 · ICLR 2022 4 · Alberto Gil Couto Pimentel Ramos, Abhinav Mehrotra, Nicholas Donald Lane, Sourav Bhattacharya

Conditional neural networks play an important role in a number of sequence-to-sequence modeling tasks, including personalized sound enhancement (PSE), speaker dependent automatic speech recognition (ASR), and generative …

Automatic Speech RecognitionAutomatic Speech Recognition (ASR)speech-recognitionSpeech Recognition+4

Training Language Models to Explain Their Own Computations

2025-11-11 · Belinda Z. Li, Zifan Carl Guo, Vincent Huang, Jacob Steinhardt 외 arxiv

Can language models (LMs) learn to faithfully describe their internal computations? Are they better able to describe themselves than other models? We study the extent to which LMs' privileged access to their own internal…