Ashok Cutkosky

I am an assistant professor at Boston University in the ECE department. I am currently on leave at Meta Fundamental AI Research until the 2027-2028 academic year. Previously, I was a research scientist at Google. I earned a PhD in computer science at Stanford University in 2018 under the supervision of Kwabena Boahen, and a AB in mathematics from Harvard in 2013. I am also a master of medicine. I'm currently excited about optimization algorithms for machine learning. I have recently worked on non-convex optimization as well as adaptive online learning. My email address is ashok (at) cutkosky (dot) com.

Selected Publications

• The Road Less Scheduled, Aaron Defazio, Xingyu Alice Yang, Harsh Mehta, Konstantin Mishchenko, Ahmed Khaled, Ashok Cutkosky. Neural Information Processinng Systems (NeurIPS), 2024.
• Optimal Stochastic Non-smooth Non-convex Optimization through Online-to-Non-convex Conversion, Ashok Cutkosky, Harsh Mehta and Francesco Orabona. International Conference on Machine Learning (ICML), 2023.
• Mechanic: A Learning Rate Tuner, Ashok Cutkosky, Aaron Defazio and Harsh Mehta. Neural Information Processsing Systems (NeurIPS), 2023.
• Momentum-Based Variance Reduction in Non-Convex SGD, Ashok Cutkosky and Francesco Orabona. Neural Information Processsing Systems (NeurIPS), 2019.
• Anytime Online-to-Batch, Optimism, and Acceleration, Ashok Cutkosky. International Conference on Machine Learning (ICML), 2019.
• Black-Box Reductions for Parameter-free Online Learning in Banach Spaces, Ashok Cutkosky and Francesco Orabona, Conference on Learning Theory (COLT), 2018

Students (PhD)

• Zhiyu Zhang
• Andrew Jacobsen
• Hoang Tran
• Qinzi Zhang
• Jiujia Zhang

Students (undergraduate)

• Fangrui Huang
• Rana Boustany
• Michelle Zyman
• Vance Raiti
• Peter Gu
• Elly Wang

Teaching

• BU EC700 Topics in Optimization Theory
• BU EC414 Introduction to Machine Learning
• BU EC525 Optimization for Machine Learning (website)
• BU EC524 Deep Learning (co-taught with Brian Kulis)

All Publications

In 2023 I achieved my stretch-goal for academic productivity: I became too lazy to keep this list fully up-to-date. If it looks out of date, please check out my google scholar profile. Try sorting by year rather than citation to see what I've been working on recently.

• Mechanic: a Learning Rate Tuner, Ashok Cutkosky, Aaron Defazio, Harsh Mehta. Neural Information Processing Systems (NeurIPS) 2023.
• Alternation makes the adversary weaker in two-player games, Volkan Cevher, Ashok Cutkosky, Ali Kavis, Georgios Piliouras, Stratis Skoulakis, Luca Viano. Neural Information Processing Systems (NeurIPS) 2023.
• Unconstrained dynamic regret via sparse coding, Zhiyu Zhang, Ashok Cutkosky, Ioannis Ch Paschalidis. Neural Information Processing Systems (NeurIPS) 2023.
• Bandit Online Linear Optimization with Hints and Queries, Aditya Bhaskara, Ashok Cutkosky, Ravi Kumar, Manish Purohit. International Conference on Machine Learning (ICML) 2023.
• Unconstrained Online Learning with Unbounded Losses, Andrew Jacobsen, Ashok Cutkosky. International Conference on Machine Learning (ICML) 2023.
• Optimal Stochastic Non-smooth Non-convex Optimization through Online-to-Non-convex Conversion, Ashok Cutkosky, Harsh Mehta, Francesco Orabona. International Conference on Machine Learning (ICML) 2023.
• Blackbox optimization of unimodal functions, Ashok Cutkosky, Abhimanyu Das, Weihao Kong, Chansoo Lee, Rajat Sen. Uncertainty in Artificial Intelligence (UAI) 2023.
• Optimal Comparator Adaptive Online Learning with Switching Cost, Zhiyu Zhang, Ashok Cutkosky, Yannis Paschalidis. Neural Information Processing Systems (NeurIPS) 2022.
• Parameter-free regret in high probability with heavy tails, Jiujia Zhang, Ashok Cutkosky. Neural Information Processing Systems (NeurIPS) 2022.
• Momentum aggregation for private non-convex erm, Hoang Tran, Ashok Cutkosky. International Conference on Machine Learning (ICML) 2023.
• Better sgd using second-order momentum, Hoang Tran, Ashok Cutkosky. Neural Information Processing Systems (NeurIPS) 2022.
• Differentially Private Online-to-Batch for Smooth Losses, Qinzi Zhang, Hoang Tran, Ashok Cutkosky. Neural Information Processing Systems (NeurIPS) 2022.
• PDE-Based Optimal Strategy for Unconstrained Online Learning, Zhiyu Zhang, Ashok Cutkosky, and Yannis Paschalidis. International Conference on Machine Learning (ICML) 2022.
• Parameter-Free Mirror Descent, Andrew Jacobsen and Ashok Cutkosky. Conference on Learning Theory (COLT) 2022.
• Leveraging Initial Hints for Free in Stochastic Linear Bandits, Ashok Cutkosky, Chris Dann, Abhimanyu Das, Qiuyi (Richard) Zhang. International Conference on Algorithmic Learning Theory (ALT) 2022.
• Implicit Parameter-free Online Learning with Truncated Linear Models, Keyi Chen, Ashok Cutkosky, Francesco Orabona. International Conference on Algorithmic Learning Theory (ALT) 2022.
• Adversarial Tracking Control via Strongly Adaptive Online Learning with Memory, Zhiyu Zhang, Ashok Cutkosky, Yannis Paschalidis. International Conference on Artificial Intillgence and Statistics (AISTATS) 2022.
• High-probability Bounds for Non-Convex Stochastic Optimization with Heavy Tails, Ashok Cutkosky, Harsh Mehta. Neural Information Processing Systems (NeurIPS) 2021. [oral]
• Logarithmic Regret from Sublinear Hints, Aditya Bhaskara, Ashok Cutkosky, Ravi Kumar, Manish Purohit. Neural Information Processing Systems (NeurIPS) 2021.
• Online Selective Classification with Limited Feedback, Aditya Gangrade, Anil Kag, Ashok Cutkosky, Venkatesh Saligrama. Neural Information Processing Systems (NeurIPS) 2021. [spotlight]
• Dynamic Balancing for Model Selection in Bandits and RL, Ashok Cutkosky, Christoph Dann, Abhimanyu Das, Claudio Gentile, Aldo Pacchiano, Manish Purohit. International Conference on Machine Learning (ICML) 2021.
• Robust Pure Exploration in Linear Bandits with Limited Budget, Ayya Alieva, Ashok Cutkosky, Abhimanyu Das. International Conference on Machine Learning (ICML) 2021.
• Power of Hints for Online Learning with Movement Costs, Aditya Bhaskara, Ashok Cutkosky, Ravi Kumar, Manish Purohit. International Conference on Artificial Intillgence and Statistics (AISTATS) 2021.
• Extreme Memorization via Scale of Initialization. Harsh Mehta, Ashok Cutkosky, Benham Neyshabur. International Conference on Learning Representations (ICLR) 2021.
• Online Linear Optimization with Many Hints, Aditya Bhaskara, Ashok Cutkosky, Ravi Kumar, Manish Purohit. Advances Neural Information Processing Systems (NeurIPS), 2020.
• Comparator-Adaptive Convex Bandits, Dirk van der Hoeven, Ashok Cutkosky, Haipeng Luo. Advances in Neural Information Processing Systems (NeurIPS), 2020.
• Better Full-Matrix Regret via Parameter-free Online Learning, Ashok Cutkosky. Advances in Neural Information Processing Systems (NeurIPS), 2020.
• Momentum Improves Normalized SGD, Ashok Cutkosky and Harsh Mehta. International Conference on Machine Learning (ICML), 2020.
• Online Learning with Imperfect Hints, Aditya Bhaskara, Ashok Cutkosky, Ravi Kumar, and Manish Purohit. International Conference on Machine Learning (ICML), 2020
• Parameter-Free, Dynamic, and Strongly-Adaptive Online Learning, Ashok Cutkosky. International Conference on Machine Learning (ICML), 2020
• Momentum-Based Variance Reduction in Non-Convex SGD, Ashok Cutkosky and Francesco Orabona. Neural Information Processsing Systems (NeurIPS), 2019.
• Kernel Truncated Randomized Ridge Regression: Optimal Rates and Low Noise Acceleration, Kwang-Sung Jun, Ashok Cutkosky, Francesco Orabona. Neural Information Processing Systems (NeurIPS), 2019.
• Matrix-Free Preconditioning in Online Learning, Ashok Cutkosky and Tamas Sarlos. International Conference on Machine Learning (ICML), 2019. [code] [long talk]
• Anytime Online-to-Batch, Optimism, and Acceleration, Ashok Cutkosky. International Conference on Machine Learning (ICML), 2019. [long talk]
• Surrogate Losses for Online Learning of Stepsizes in Stochastic Non-Convex Optimization, Zhenxun Zhuang, Ashok Cutkosky, and Francesco Orabona. International Conference on Machine Learning (ICML), 2019.
• Artificial Constraints and Hints for Unbounded Online Learning, Ashok Cutkosky. Conference on Learning Theory (COLT), 2019.
• Combining Online Learning Guarantees, Ashok Cutkosky. Conference on Learning Theory (COLT), 2019.
• Distributed Stochastic Optimization via Adaptive Stochastic Gradient Descent, Ashok Cutkosky and Robert Busa-Fekete. Advances in Neural Information Processing Systems (NeurIPS) 2018.
• Black-Box Reductions for Parameter-free Online Learning in Banach Spaces, Ashok Cutkosky and Francesco Orabona, Conference on Learning Theory (COLT), 2018
• Stochastic and Adversarial Online Learning Without Hyperparameters, Ashok Cutkosky and Kwabena Boahen, Advances in Neural Information Processing Systems (NIPS), 2017.
• Online Learning Without Prior Information, Ashok Cutkosky and Kwabena Boahen, Conference on Learning Theory (COLT), 2017 [Best Student Paper Award]. [code]
• Online Convex Optimization with Unconstrained Domains and Losses, Ashok Cutkosky and Kwabena Boahen, Advances in Neural Information Processing Systems (NIPS), 2016. [magical video]
• Bloom Features, Ashok Cutkosky and Kwabena Boahen, IEEE International Conference on Computation Science and Computational Intelligence, 2015.
• Chromatin extrusion explains key features of loop and domain formation in wild-type and engineered genomes, Adrian Sanborn et al, Proceedings of the National Academy of Sciences, 2015.

Other

• PhD dissertation: Algorithms and Lower Bounds for Parameter-Free Online Learning, May 2018
• Undergrad Thesis: Polymer Simulations and DNA Topology, Advised by Erez Lieberman Aiden. [Hoopes Prize and Captain Jonathan Fay Prize winner]
• Associated Primes of the Square of the Alexander Dual of Hypergraphs, high school math project. Advised by Chris Francisco.