Gintare Karolina Dziugaite

Publications

The Dynamics of Functional Diversity throughout Neural Network Training. Lee Zamparo, Marc-Etienne Brunet , Thomas George, Sepideh Kharaghani, Gintare Karolina Dziugaite. At Conference on Neural Information Processing Systems (NeurIPS), 2021.

Pruning Neural Networks at Initialization: Why are We Missing the Mark?. Jonathan Frankle, Gintare Karolina Dziugaite, Daniel M. Roy, Michael Carbin. At International Conference on Learning Representations (ICLR), 2021.

On the role of data in PAC-Bayes bounds. Gintare Karolina Dziugaite, Kyle Hsu, Waseem Gharbieh, Gabriel Arpino, Daniel M. Roy. At International Conference on Artificial Intelligence and Statistics (AISTATS), 2021.

An empirical study of loss landscape geometry and evolution of the data-dependent Neural Tangent Kernel. Stanislav Fort, Gintare Karolina Dziugaite, Mansheej Paul, Sepideh Kharaghani, Daniel M. Roy, Surya Ganguli. At Conference on Neural Information Processing Systems (NeurIPS), 2020.

In search of robust measures of generalization. Gintare Karolina Dziugaite, Alexandre Drouin, Brayden (Brady) Neal, Nitarshan Rajkumar, Ethan Victor Caballero, Linbo Wang, Ioannis Mitliagkas, Daniel M. Roy. At Conference on Neural Information Processing Systems (NeurIPS), 2020.

PDF Cite Code Video

Pruning Neural Networks at Initialization: Why Are We Missing the Mark? . Jonathan Frankle, Gintare Karolina Dziugaite, Daniel M. Roy, Michael Carbin. At Workshop at the Neural Information Processing Systems (NeurIPS), 2020.

Sharpened Generalization Bounds based on Conditional Mutual Information and an Application to Noisy-Gradient Iterative Algorithms. Mahdi Haghifam, Jeffrey Negrea, Ashish Khisti, Daniel M. Roy, Gintare Karolina Dziugaite. At Conference on Neural Information Processing Systems (NeurIPS), 2020.

On the Information Complexity of Proper Learners for VC Classes in the Realizable Case. Mahdi Haghifam, Gintare Karolina Dziugaite, Shay Moran, Daniel M. Roy. At ArXiv, 2020.

Enforcing Interpretability and its Statistical Impacts: Trade-offs between Accuracy and Interpretability. Gintare Karolina Dziugaite, Shai Ben-David, Daniel M. Roy. At ArXiv, 2020.

In Defense of Uniform Convergence: Generalization via derandomization with an application to interpolating predictors. Jeffrey Negrea, Gintare Karolina Dziugaite, Daniel M. Roy. At International Conference on Machine Learning (ICML), 2020.

Linear Mode Connectivity and the Lottery Ticket Hypothesis. Jonathan Frankle, Gintare Karolina Dziugaite, Daniel M. Roy, Michael Carbin. At International Conference on Machine Learning (ICML), 2020.

RelatIF: Identifying Explanatory Training Examples via Relative Influence. Elnaz Barshan, Marc-Etienne Brunet , Gintare Karolina Dziugaite. At International Conference on Artificial Intelligence and Statistics (AISTATS), 2020.

Stochastic Neural Network with Kronecker Flow. Chin-Wei Huang, Ahmed Touati, Pascal Vincent, Gintare Karolina Dziugaite, Alexandre Lacoste, Aaron Courville. At International Conference on Artificial Intelligence and Statistics (AISTATS), 2020.

Stabilizing the Lottery Ticket Hypothesis. Jonathan Frankle, Gintare Karolina Dziugaite, Daniel M. Roy, Michael Carbin. At Association for the Advancement of Artificial Intelligence (AAAI), 2020.

Linear Mode Connectivity and the Lottery Ticket Hypothesis. Jonathan Frankle, Gintare Karolina Dziugaite, Daniel M. Roy, Michael Carbin. At Women in Machine Learning (WiML), 2019.

Optimal PAC-Bayes bounds require distribution- and data-dependent priors. Kyle Hsu, Waseem Gharbieh, Daniel M. Roy, Gintare Karolina Dziugaite. At Women in Machine Learning (WiML), 2019.

Information-Theoretic Generalization Bounds for SGLD via Data-Dependent Estimates. Jeffrey Negrea, Mahdi Haghifam, Gintare Karolina Dziugaite, Ashish Khisti, Daniel M. Roy. At Conference on Neural Information Processing Systems (NeurIPS), 2019.

Optimal PAC-Bayes bounds require distribution- and data-dependent priors. Kyle Hsu, Waseem Gharbieh, Daniel M. Roy, Gintare Karolina Dziugaite. At Workshop at the Neural Information Processing Systems (NeurIPS), 2019.

Towards generalization guarantees for SGD: Data-dependent PAC-Bayes priors. Gabriel Arpino, Daniel M. Roy, Gintare Karolina Dziugaite. At Workshop at the Neural Information Processing Systems (NeurIPS), 2019.

Information-Theoretic Generalization Bounds for SGLD via Data-Dependent Estimates. Jeffrey Negrea, Mahdi Haghifam, Gintare Karolina Dziugaite, Ashish Khisti, Daniel M. Roy. At Workshop at the International Conference on Machine Learning (ICML), 2019.

Stochastic Neural Network with Kronecker Flow. Chin-Wei Huang, Ahmed Touati, Pascal Vincent, Gintare Karolina Dziugaite, Alexandre Lacoste, Aaron Courville. At Workshop at the International Conference on Machine Learning (ICML), 2019.

The Lottery Ticket Hypothesis at Scale. Jonathan Frankle, Gintare Karolina Dziugaite, Daniel M. Roy, Michael Carbin. At ArXiv, 2019.

An escape-time analysis of SGD. Mufan Li, Gintare Karolina Dziugaite, Philippe Casgrain. At Women in Machine Learning (WiML), 2018.

Data-dependent PAC-Bayes priors via differential privacy. Gintare Karolina Dziugaite, Daniel M. Roy. At Conference on Neural Information Processing Systems (NeurIPS), 2018.

Tighter Nonvacuous PAC-Bayes Bounds: Beyond Stability of SGD. Gabriel Arpino, Gintare Karolina Dziugaite, Daniel M. Roy. At Workshop at the Neural Information Processing Systems (NeurIPS), 2018.