Navid Azizan

Foundations of Optimization in Deep Learning

• Mirror Descent Maximizes Generalized Margin and Can Be Implemented Efficiently [pdf] [cite]
  H. Sun, K. Ahn, C. Thrampoulidis, and N. Azizan
  Advances in Neural Information Processing Systems (NeurIPS), 2022

• Beyond Implicit Regularization: Avoiding Overfitting via Regularizer Mirror Descent [pdf] [cite]
  N. Azizan, S. Lale, and B. Hassibi
  International Conference on Machine Learning (ICML) Overparameterization Workshop, 2021
  Selected for spotlight presentation

• Stochastic Mirror Descent on Overparameterized Nonlinear Models [pdf] [cite]
  N. Azizan, S. Lale, and B. Hassibi
  IEEE Transactions on Neural Networks and Learning Systems, 2021
  International Conference on Machine Learning (ICML) Generalization Workshop, 2019

• Stochastic Gradient/Mirror Descent: Minimax Optimality and Implicit Regularization [pdf] [cite]
  N. Azizan and B. Hassibi
  International Conference on Learning Representations (ICLR), 2019
  Neural Information Processing Systems (NeurIPS) Deep Learning Theory Workshop, 2018
  Winner of the ITA Graduation-Day Gold Award, 2020

Safety Assurances for Machine Learning

• Uncertainty-Aware Meta-Learning for Multimodal Task Distributions [pdf] [cite]
  C. Almecija, A. Sharma, Y. Park, and N. Azizan
  Neural Information Processing Systems (NeurIPS) Meta-Learning Workshop, 2022

• Embedding Reliability: On the Predictability of Downstream Performance [pdf] [cite]
  S. Ardeshir and N. Azizan
  Neural Information Processing Systems (NeurIPS) ML Safety Workshop, 2022

• Sketching Curvature for Efficient Out-of-Distribution Detection for Deep Neural Networks [pdf] [cite]
  A. Sharma, N. Azizan, and M. Pavone
  Uncertainty in Artificial Intelligence (UAI), 2021
  Selected for full oral presentation

• A Unified View of SDP-based Neural Network Verification through Completely Positive Programming [pdf] [cite]
  R. Brown, E. Schmerling, N. Azizan, M. Pavone
  Artificial Intelligence and Statistics (AISTATS) 2022

Continual and Lifelong Learning

• ORFit: One-Pass Learning via Bridging Orthogonal Gradient Descent and Recursive Least-Squares [pdf] [cite]
  Y. Min, K. Ahn, and N. Azizan
  Proceedings of the IEEE Conference on Decision and Control (CDC), 2022 (Invited Paper)

• Orthogonal Gradient Descent for Continual Learning [pdf] [cite]
  M. Farajtabar, N. Azizan, A. Mott, and A. Li
  International Conference on Artificial Intelligence and Statistics (AISTATS), 2020

Learning for Control and Decision Making

• Learning Control-Oriented Dynamical Structure from Data [pdf] [cite]
  S. M. Richards, J.-J. E. Slotine, N. Azizan, and M. Pavone
  preprint, 2023

• Control-Oriented Meta-Learning [pdf] [conference version] [cite]
  S. M. Richards, N. Azizan, J.-J. E. Slotine, and M. Pavone
  International Journal of Robotics Research (IJRR), 2023
  Robotics: Science and Systems (RSS), 2021

• Online Learning for Traffic Routing under Unknown Preferences [pdf] [cite]
  D. Jalota, K. Gopalakrishnan, N. Azizan, R. Johari, and M. Pavone
  International Conference on Artificial Intelligence and Statistics (AISTATS), 2023

Markets for the Smart Grid

• Optimal Pricing in Markets with Non-Convex Costs [pdf] [cite]
  N. Azizan, Y. Su, K. Dvijotham, and A. Wierman
  Operations Research (OR), 2020
  ACM Conference on Economics and Computation (EC), 2019

• Opportunities for Price Manipulation by Aggregators in Electricity Markets [pdf] [short version] [cite]
  N. Azizan, K. Dvijotham, N. Chen, and A. Wierman
  IEEE Transactions on Smart Grid (TSG), 2017
  ACM Performance Evaluation Review (PER), 2016
  Winner of the ACM GREENMETRICS Best Student Paper Award, 2016

Federated/Distributed Computation

• Distributed Solution of Large-Scale Linear Systems via Accelerated Projection-Based Consensus [pdf] [cite]
  N. Azizan, F. Lahouti, S. Avestimehr, and B. Hassibi
  IEEE Transactions on Signal Processing (TSP), 2019
  Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2018
  Winner of the Amazon Fellowship, 2017

• Improving Distributed Gradient Descent Using Reed-Solomon Codes [pdf] [cite]
  W. Halbawi, N. Azizan, F. Salehi, and B. Hassibi
  Submitted to the IEEE Transactions on Information Theory
  Proceedings of the IEEE International Symposium on Information Theory (ISIT), 2018

Stochastic Optimization Algorithms in Estimation

• A Stochastic Interpretation of Stochastic Mirror Descent: Risk-Sensitive Optimality [pdf] [cite]
  N. Azizan and B. Hassibi
  Proceedings of the IEEE Conference on Decision and Control (CDC), 2019

• A Characterization of Stochastic Mirror Descent Algorithms and their Convergence Properties [pdf] [cite]
  N. Azizan and B. Hassibi
  Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2019

Epidemics in Complex Networks

• Improved Bounds on the Epidemic Threshold of Exact SIS Models on Complex Networks [pdf] [cite]
  N. Azizan, C. Thrampoulidis, and B. Hassibi
  Proceedings of the IEEE Conference on Decision and Control (CDC), 2016 (Invited Paper)

• Analysis of Exact and Approximated Epidemic Models over Complex Networks [pdf] [cite]
  N. Azizan, H. J. Ahn, and B. Hassibi
  arXiv:1609.09565

• SIRS Epidemics on Complex Networks: Concurrence of Exact Markov Chain and Approximated Models [pdf] [cite]
  N. Azizan and B. Hassibi
  Proceedings of the IEEE Conference on Decision and Control (CDC), 2015 (Invited Paper)