Fetching the paper…

Reading the bibliography…

2021

Robust Semantic Interpretability: Revisiting Concept Activation Vectors

Pfau, Jacob, Young, Albert T., Wei, Jerome et al.

Original
PDF
Post on X

Understand

Interpretability methods for image classification assess model trustworthiness by attempting to expose whether the model is systematically biased or attending to the same cues as a human would.

Saliency methods for feature attribution dominate the interpretability literature, but these methods do not address semantic concepts such as the textures, colors, or genders of objects within an image.
Our proposed Robust Concept Activation Vectors (RCAV) quantifies the effects of semantic concepts on individual model predictions and on model behavior as a whole.
RCAV calculates a concept gradient and takes a gradient ascent step to assess model sensitivity to the given concept.

Robust Semantic Interpretability: Revisiting Concept Activation Vectors · Around

Built on

“Identifying and Correcting Label Bias in Machine Learning”, 2019
Original
Heinrich Jiang and Ofir Nachum · 1901
Earlier work this paper cites.
“Fooling Neural Network Interpretations via Adversarial Model Manipulation”, 2019
Original
Juyeon Heo, Sunghwan Joo and Taesup Moon · 1902
Earlier work this paper cites.
“Benchmarking Attribution Methods with Relative Feature Importance”, 2019
Original
Mengjiao Yang and Been Kim · 1907
Earlier work this paper cites.
“Neural Additive Models: Interpretable Machine Learning with Neural Nets”, 2020
Original
Rishabh Agarwal et al · 2004
Earlier work this paper cites.
“Imagenet: A large-scale hierarchical image database”
J Deng et al · 2009
Earlier work this paper cites.
“Going Deeper with Convolutions”, 2014
Original
Christian Szegedy et al · 2014
Earlier work this paper cites.
“Rethinking the Inception Architecture for Computer Vision”, 2015
Original
Christian Szegedy et al · 2015
Earlier work this paper cites.
“Camelyon16 ISBI grand challenge”, 2016
Babak Bejnordi et al · 2016
Earlier work this paper cites.
“Network dissection: Quantifying interpretability of deep visual representations”
David Bau et al · 2017
Earlier work this paper cites.
“What do we need to build explainable AI systems for the medical domain?”
Andreas Holzinger, Chris Biemann, Constantinos Pattichis and Douglas Kell · 2017
Earlier work this paper cites.
“Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)”, 2017
Original
Been Kim et al · 2017
Earlier work this paper cites.

Similar

“Counterfactual Fairness”
Matt Kusner, Joshua Loftus, Chris Russell and Ricardo Silva · 2017
Cited alongside, same era.
“Right for the Right Reasons: Training Differentiable Models by Constraining their Explanations”, 2017
Original
Andrew Ross, Michael Hughes and Finale Doshi-Velez · 2017
Cited alongside, same era.
“Evaluating the Visualization of What a Deep Neural Network Has Learned”
Wojciech Samek et al · 2017
Cited alongside, same era.
“Grad-cam: Visual explanations from deep networks via gradient-based localization”
Ramprasaath Selvaraju et al · 2017
Cited alongside, same era.
“Transparency: Motivations and Challenges”, 2017
Original
Adrian Weller · 2017
Cited alongside, same era.

Then

“Human-in-the-Loop Interpretability Prior”, 2018
Original
Isaac Lage et al · 2018
Later among the works it cites.
“Confounding variables can degrade generalization performance of radiological deep learning models”, 2018
Original
John Zech et al · 2018
Later among the works it cites.
“Variable generalization performance of a deep learning model to detect pneumonia in chest radiographs: A cross-sectional study”
John Zech et al · 2018
Later among the works it cites.
“Interpretable basis decomposition for visual explanation”
Bolei Zhou, Yiyou Sun, David Bau and Antonio Torralba · 2018
Later among the works it cites.
“Automating interpretability: Discovering and testing visual concepts learned by neural networks”
Amirata Ghorbani, James Wexler and Been Kim · 2019
Later among the works it cites.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…

“Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms”

H Xiao, K Rasul and R Vollgraf · 2017

Cited alongside, same era.

“mixup: Beyond Empirical Risk Minimization”, 2017

Hongyi Zhang, Moustapha Cisse, Yann Dauphin and David Lopez-Paz · 2017

Cited alongside, same era.

“Sanity Checks for Saliency Maps”

Julius Adebayo et al · 2018

Cited alongside, same era.

“Discovering Interpretable Representations for Both Deep Generative and Discriminative Models”

Tameem Adel, Zoubin Ghahramani and Adrian Weller · 2018

Cited alongside, same era.

“Path-Specific Counterfactual Fairness”, 2018

Silvia Chiappa and Thomas Gillam · 2018

Cited alongside, same era.

“Empirical Risk Minimization Under Fairness Constraints”

Michele Donini et al · 2018

Cited alongside, same era.

“Mixup as locally linear out-of-manifold regularization”

Hongyu Guo, Yongyi Mao and Richong Zhang · 2019

Later among the works it cites.

“A Benchmark for Interpretability Methods in Deep Neural Networks”

Sara Hooker, Dumitru Erhan, Pieter-Jan Kindermans and Been Kim · 2019

Later among the works it cites.

“Learning Not to Learn: Training Deep Neural Networks With Biased Data”

Byungju Kim et al · 2019

Later among the works it cites.

“Repair: Removing representation bias by dataset resampling”

Yi Li and Nuno Vasconcelos · 2019

Later among the works it cites.

“Definitions, methods, and applications in interpretable machine learning”

W Murdoch et al · 2019

Later among the works it cites.

“Interpreting cnns via decision trees”

Quanshi Zhang, Yu Yang, Haotian Ma and Ying Wu · 2019

Later among the works it cites.