no code implementations • 12 Feb 2024 • Andrew Bai, Chih-Kuan Yeh, Cho-Jui Hsieh, Ankur Taly
We propose a novel sampling scheme, mix-cd, that identifies and prioritizes samples that actually face forgetting, which we call collateral damage.
no code implementations • 28 Aug 2023 • Clark Barrett, Brad Boyd, Elie Burzstein, Nicholas Carlini, Brad Chen, Jihye Choi, Amrita Roy Chowdhury, Mihai Christodorescu, Anupam Datta, Soheil Feizi, Kathleen Fisher, Tatsunori Hashimoto, Dan Hendrycks, Somesh Jha, Daniel Kang, Florian Kerschbaum, Eric Mitchell, John Mitchell, Zulfikar Ramzan, Khawaja Shams, Dawn Song, Ankur Taly, Diyi Yang
However, GenAI can be used just as well by attackers to generate new attacks and increase the velocity and efficacy of existing attacks.
no code implementations • 5 Jun 2022 • Aya Abdelsalam Ismail, Sercan Ö. Arik, Jinsung Yoon, Ankur Taly, Soheil Feizi, Tomas Pfister
In addition to constituting a standalone inherently-interpretable architecture, IME has the premise of being integrated with existing DNNs to offer interpretability to a subset of samples while maintaining the accuracy of the DNNs.
1 code implementation • 24 Feb 2022 • Chih-Kuan Yeh, Ankur Taly, Mukund Sundararajan, Frederick Liu, Pradeep Ravikumar
However, we observe that since the activation connected to the last layer of weights contains "shared logic", the data influenced calculated via the last layer weights prone to a ``cancellation effect'', where the data influence of different examples have large magnitude that contradicts each other.
1 code implementation • 27 Mar 2021 • David Watson, Limor Gultchin, Ankur Taly, Luciano Floridi
Necessity and sufficiency are the building blocks of all successful explanations.
Explainable artificial intelligence Explainable Artificial Intelligence (XAI)
1 code implementation • 17 Sep 2019 • Luke Merrick, Ankur Taly
While existing papers focus on the axiomatic motivation of Shapley values, and efficient techniques for computing them, they offer little justification for the game formulations used, and do not address the uncertainty implicit in their methods' outputs.
no code implementations • 13 Sep 2019 • Umang Bhatt, Alice Xiang, Shubham Sharma, Adrian Weller, Ankur Taly, Yunhan Jia, Joydeep Ghosh, Ruchir Puri, José M. F. Moura, Peter Eckersley
Yet there is little understanding of how organizations use these methods in practice.
1 code implementation • 29 Apr 2019 • Divya Gopinath, Hayes Converse, Corina S. Pasareanu, Ankur Taly
We present techniques for automatically inferring formal properties of feed-forward neural networks.
no code implementations • 27 Nov 2018 • Kevin McCloskey, Ankur Taly, Federico Monti, Michael P. Brenner, Lucy Colwell
The dataset bias makes these models unreliable for accurately revealing information about the mechanisms of protein-ligand binding.
no code implementations • 27 Sep 2018 • Sahaj Garg, Vincent Perot, Nicole Limtiaco, Ankur Taly, Ed H. Chi, Alex Beutel
In this paper, we study counterfactual fairness in text classification, which asks the question: How would the prediction change if the sensitive attribute referenced in the example were different?
no code implementations • 11 Jun 2018 • Mukund Sundararajan, Ankur Taly
Local explanation methods, also known as attribution methods, attribute a deep network's prediction to its input (cf.
4 code implementations • ACL 2018 • Pramod Kaushik Mudrakarta, Ankur Taly, Mukund Sundararajan, Kedar Dhamdhere
Our strongest attacks drop the accuracy of a visual question answering model from $61. 1\%$ to $19\%$, and that of a tabular question answering model from $33. 5\%$ to $3. 3\%$.
no code implementations • 12 Mar 2018 • Pramod Kaushik Mudrakarta, Ankur Taly, Mukund Sundararajan, Kedar Dhamdhere
The large impact on the performance of the KDG model suggests that the pruning may be a useful pre-processing step in training other semantic parsers as well.
no code implementations • 10 Sep 2017 • Kedar Dhamdhere, Kevin S. McCurley, Mukund Sundararajan, Ankur Taly
We study question-answering over semi-structured data.
32 code implementations • ICML 2017 • Mukund Sundararajan, Ankur Taly, Qiqi Yan
We study the problem of attributing the prediction of a deep network to its input features, a problem previously studied by several other works.
Explainable artificial intelligence Interpretable Machine Learning
no code implementations • 8 Nov 2016 • Mukund Sundararajan, Ankur Taly, Qiqi Yan
Unfortunately, in nonlinear deep networks, not only individual neurons but also the whole network can saturate, and as a result an important input feature can have a tiny gradient.