\n
\n\n \n \n \n \n \n \n Adversarially Robust Models may not Transfer Better: Sufficient Conditions for Domain Transferability from the View of Regularization.\n \n \n \n \n\n\n \n Xu, X.; Zhang, J. Y; Ma, E.; Son, H. H.; Koyejo, S.; and Li, B.\n\n\n \n\n\n\n In Chaudhuri, K.; Jegelka, S.; Song, L.; Szepesvari, C.; Niu, G.; and Sabato, S., editor(s),
Proceedings of the 39th International Conference on Machine Learning, volume 162, of
Proceedings of Machine Learning Research, pages 24770–24802, 17–23 Jul 2022. PMLR\n
\n\n
\n\n
\n\n
\n\n \n \n Paper\n \n \n\n \n\n \n link\n \n \n\n bibtex\n \n\n \n \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@InProceedings{pmlr-v162-xu22n,\n title = \t {Adversarially Robust Models may not Transfer Better: Sufficient Conditions for Domain Transferability from the View of Regularization},\n author = {Xu, Xiaojun and Zhang, Jacky Y and Ma, Evelyn and Son, Hyun Ho and Koyejo, Sanmi and Li, Bo},\n booktitle = \t {Proceedings of the 39th International Conference on Machine Learning},\n pages = \t {24770--24802},\n year = \t {2022},\n editor = \t {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan},\n volume = \t {162},\n series = \t {Proceedings of Machine Learning Research},\n month = \t {17--23 Jul},\n publisher = {PMLR},\n pdf = \t {https://proceedings.mlr.press/v162/xu22n/xu22n.pdf},\n url = \t {https://proceedings.mlr.press/v162/xu22n.html},\n abstract = \t {Machine learning (ML) robustness and domain generalization are fundamentally correlated: they essentially concern data distribution shifts under adversarial and natural settings, respectively. On one hand, recent studies show that more robust (adversarially trained) models are more generalizable. On the other hand, there is a lack of theoretical understanding of their fundamental connections. In this paper, we explore the relationship between regularization and domain transferability considering different factors such as norm regularization and data augmentations (DA). We propose a general theoretical framework proving that factors involving the model function class regularization are sufficient conditions for relative domain transferability. Our analysis implies that “robustness" is neither necessary nor sufficient for transferability; rather, regularization is a more fundamental perspective for understanding domain transferability. We then discuss popular DA protocols (including adversarial training) and show when they can be viewed as the function class regularization under certain conditions and therefore improve generalization. We conduct extensive experiments to verify our theoretical findings and show several counterexamples where robustness and generalization are negatively correlated on different datasets.}\n}\n\n\n\n% for ION NAVIGATION, noticing: Only first letter of paper titles is capitalized, use capitalization for all words in journal names\n% Transactions kept in full, Proceedings became "Proc." ?\n\n% C106 \n% Shubh Gupta, Ashwin V. Kanhere, Akshay Shetty, and Grace Gao, Designing Deep Neural Networks for Sequential GNSS Positioning, Proceedings of the Institute of Navigation GNSS+ conference (ION GNSS+ 2022), Denver, CO, Sep 2022\n
\n
\n\n\n
\n Machine learning (ML) robustness and domain generalization are fundamentally correlated: they essentially concern data distribution shifts under adversarial and natural settings, respectively. On one hand, recent studies show that more robust (adversarially trained) models are more generalizable. On the other hand, there is a lack of theoretical understanding of their fundamental connections. In this paper, we explore the relationship between regularization and domain transferability considering different factors such as norm regularization and data augmentations (DA). We propose a general theoretical framework proving that factors involving the model function class regularization are sufficient conditions for relative domain transferability. Our analysis implies that “robustness\" is neither necessary nor sufficient for transferability; rather, regularization is a more fundamental perspective for understanding domain transferability. We then discuss popular DA protocols (including adversarial training) and show when they can be viewed as the function class regularization under certain conditions and therefore improve generalization. We conduct extensive experiments to verify our theoretical findings and show several counterexamples where robustness and generalization are negatively correlated on different datasets.\n
\n\n\n
\n
\n\n \n \n \n \n \n \n Designing Deep Neural Networks for Sequential GNSS Positioning.\n \n \n \n \n\n\n \n Gupta, S.; Kanhere, A. V; Shetty, A.; and Gao, G.\n\n\n \n\n\n\n In
Proceedings of the 35th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2022), pages 1209–1219, 2022. \n
\n\n
\n\n
\n\n
\n\n \n \n Paper\n \n \n\n \n\n \n link\n \n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@inproceedings{gupta2022designing,\n title={Designing Deep Neural Networks for Sequential {GNSS} Positioning},\n author={Gupta, Shubh and Kanhere, Ashwin V and Shetty, Akshay and Gao, Grace},\n booktitle={Proceedings of the 35th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2022)},\n pages={1209--1219},\n year={2022},\n url={https://doi.org/10.33012/2022.18494}\n}\n\n% C105\n% Daniel Neamati, Sriramya Bhamidipati, and Grace Gao, Set-Based Ambiguity Reduction in Shadow Matching with Iterative GNSS Pseudoranges, Proceedings of the Institute of Navigation GNSS+ conference (ION GNSS+ 2022), Denver, CO, Sep 2022.\n
\n
\n\n\n\n
\n
\n\n \n \n \n \n \n \n Set-Based Ambiguity Reduction in Shadow Matching with Iterative GNSS Pseudoranges.\n \n \n \n \n\n\n \n Neamati, D.; Bhamidipati, S.; and Gao, G.\n\n\n \n\n\n\n In
Proceedings of the 35th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2022), pages 1093–1107, 2022. \n
\n\n
\n\n
\n\n
\n\n \n \n Paper\n \n \n\n \n\n \n link\n \n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@inproceedings{neamati2022set,\n title={Set-Based Ambiguity Reduction in Shadow Matching with Iterative GNSS Pseudoranges},\n author={Neamati, Daniel and Bhamidipati, Sriramya and Gao, Grace},\n booktitle={Proceedings of the 35th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2022)},\n pages={1093--1107},\n year={2022},\n url={https://doi.org/10.33012/2022.18467}\n}\n\n% C103\n% Adyasha Mohanty and Grace Gao, Learning GNSS Positioning Corrections for Smartphones using Graph Convolution Neural Networks, Proceedings of the Institute of Navigation GNSS+ conference (ION GNSS+ 2022), Denver, CO, Sep 2022.\n
\n
\n\n\n\n
\n
\n\n \n \n \n \n \n \n Learning GNSS Positioning Corrections for Smartphones using Graph Convolution Neural Networks.\n \n \n \n \n\n\n \n Mohanty, A.; and Gao, G.\n\n\n \n\n\n\n In
Proceedings of the 35th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2022), pages 2215–2225, 2022. \n
\n\n
\n\n
\n\n
\n\n \n \n Paper\n \n \n\n \n\n \n link\n \n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@inproceedings{mohanty2022learning,\n title={Learning {GNSS} Positioning Corrections for Smartphones using Graph Convolution Neural Networks},\n author={Mohanty, Adyasha and Gao, Grace},\n booktitle={Proceedings of the 35th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2022)},\n pages={2215--2225},\n year={2022},\n url={https://doi.org/10.33012/2022.18372}\n}\n\n% C98\n% Amr Alanwar, Mahmoud Selim, Shreyas Kousik, Grace Gao, Marco Pavone and Karl H. Johansson, Safe Reinforcement Learning Using Black-Box Reachability Analysis, IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2022.\n
\n
\n\n\n\n
\n
\n\n \n \n \n \n \n Precise relative positioning via tight-coupling of GPS carrier phase and multiple UWBs.\n \n \n \n\n\n \n Mohanty, A.; Wu, A.; Bhamidipati, S.; and Gao, G.\n\n\n \n\n\n\n
IEEE Robotics and Automation Letters, 7(2): 5757–5762. 2022.\n
\n\n
\n\n
\n\n
\n\n \n\n \n \n doi\n \n \n\n \n link\n \n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@article{mohanty2022precise,\n title={Precise relative positioning via tight-coupling of {GPS} carrier phase and multiple {UWBs}},\n author={Mohanty, Adyasha and Wu, Asta and Bhamidipati, Sriramya and Gao, Grace},\n journal={IEEE Robotics and Automation Letters},\n volume={7},\n number={2},\n pages={5757--5762},\n year={2022},\n publisher={IEEE},\n doi={https://doi.org/10.1109/LRA.2022.3145051}\n}\n\n% C92\n% Sriramya Bhamidipati, Shreyas Kousik and Grace Gao, Set-valued Shadow Matching using Zonotopes, Proceedings of the Institute of Navigation GNSS+ conference (ION GNSS+ 2021), St. Louis, MO, Sep 2021.\n
\n
\n\n\n\n
\n
\n\n \n \n \n \n \n Set-Valued Shadow Matching Using Zonotopes for 3-D Map-Aided GNSS Localization.\n \n \n \n\n\n \n Bhamidipati, S.; Kousik, S.; and Gao, G.\n\n\n \n\n\n\n
Navigation: Journal of the Institute of Navigation. 2022.\n
\n\n
\n\n
\n\n
\n\n \n\n \n\n \n link\n \n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@article{bhamidipati2022set,\n title={Set-Valued Shadow Matching Using Zonotopes for {3-D} Map-Aided {GNSS} Localization},\n author={Bhamidipati, Sriramya and Kousik, Shreyas and Gao, Grace},\n journal={Navigation: Journal of the Institute of Navigation},\n year={2022}\n}\n\n% J33\n% Amr Alanwar, Mahmoud Selim, Shreyas Kousik, Grace Gao, Marco Pavone and Karl H. Johansson, Safe Reinforcement Learning Using Black-Box Reachability Analysis, IEEE Robotics and Automation Letter, 2022, vol. 7, no. 4, pp. 10665-10672, Oct. 2022, doi: 10.1109/LRA.2022.3192205.\n
\n
\n\n\n\n
\n
\n\n \n \n \n \n \n \n Safe Reinforcement Learning Using Black-Box Reachability Analysis.\n \n \n \n \n\n\n \n Selim, M.; Alanwar, A.; Kousik, S.; Gao, G.; Pavone, M.; and Johansson, K.\n\n\n \n\n\n\n
IEEE Robotics and Automation Letters, 7(4): 10665–10672. 2022.\n
\n\n
\n\n
\n\n
\n\n \n \n Paper\n \n \n\n \n\n \n link\n \n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@article{alanwar2022safe,\n title={Safe Reinforcement Learning Using Black-Box Reachability Analysis},\n author={ Selim, Mahmoud and Alanwar, Amr and Kousik, Shreyas and Gao, Grace and Pavone, Marco and Johansson, Karl},\n journal={IEEE Robotics and Automation Letters},\n volume={7},\n number={4},\n pages={10665--10672},\n year={2022},\n publisher={IEEE},\n url={https://doi.org/10.1109/LRA.2022.3192205}\n}\n\n% J32\n% Shreyas Kousik, Adam Dai and Grace Gao, Ellipsotopes: Combining Ellipsoids and Zonotopes for Reachability Analysis and Fault Detection, IEEE Transactions on Automatic Control, 2022, doi: 10.1109/TAC.2022.3191750.\n
\n
\n\n\n\n
\n
\n\n \n \n \n \n \n \n Efficient Neural Network Analysis with Sum-of-Infeasibilities.\n \n \n \n \n\n\n \n Wu, H.; Zeljić, A.; Katz, G.; and Barrett, C.\n\n\n \n\n\n\n In Fisman, D.; and Rosu, G., editor(s),
International Conference on Tools and Algorithms for the Construction and Analysis of Systems (TACAS), volume 13243, of
Lecture Notes in Computer Science, pages 143–163, April 2022. Springer\n
\n\n
\n\n
\n\n
\n\n \n \n Paper\n \n \n\n \n \n doi\n \n \n\n \n link\n \n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@inproceedings{WZK+22,\n author = {Haoze Wu and Aleksandar Zelji{\\'c} and Guy Katz and Clark\n\tBarrett},\n editor = {Dana Fisman and Grigore Rosu},\n title = {Efficient Neural Network Analysis with Sum-of-Infeasibilities},\n booktitle = tacas,\n series = {Lecture Notes in Computer Science},\n volume = {13243},\n pages = {143--163},\n publisher = {Springer},\n month = apr,\n year = {2022},\n doi = {10.1007/978-3-030-99524-9_24},\n url = {http://www.cs.stanford.edu/~barrett/pubs/WZK+22.pdf}\n}\n\n\n
\n
\n\n\n\n