Nadav Hallak
Assistant Professor
The Faculty of Data and Decision Sciences
The Technion

Contact:

Nadav Hallak,
Assistant Professor
The Faculty of Data and Decision Sciences, The Technion
Office: Bloomfield Building, room 508
Phone: +972-77-887-4415
E-mail: ndvhllk@technion.ac.il

Research Interests:

• Continuous optimization: theory and algorithms

• First order methods

• Nonconvex optimization

• Stochastic gradient with heavy-tail noise

• Online nonconvex optimization

• Portfolion optimization with Kelly criterion

• Nonconvex nonsmooth distributed optimization for power systems

• Sparse optimization

Current focus: 

• Stochastic and relative error sparse optimization

• Relative-error extended model with stochastic elements

• Acceleration with memory

Research:

For the latest list please refer to my google scholar profile

Recent and submitted:

• D. Greenstein and N. Hallak (2026). Achieving directional-stationarity from a single random direction step. Advances in Neural Information Processing Systems (NeurIPS 2026); forthcoming. [preprint] 

• D. Greenstein and N. Hallak (2026). An augmented Lagrangian approach for composite problems with random linear operator. Journal of Optimization Theory and Applications, accepted October 2026; forthcoming. [preprint] 

• N. Hallak (2025). Randomized oracle-based support exploration for constrained sparse optimization. Submitted to Open Journal of Mathematical Optimization, May 2026.  [paper] [code] 

• N. Hallak and K. Y. Levy (2026). A study of first-order methods with a probabilistic relative-error gradient oracle. Submitted to Journal of Machine Learning Research, September 2026. [paper]

• N. Hallak (2026). A stationarity-based regret framework for time-varying nonconvex composite optimization. Submitted to Mathematical Programming, February 2026.[paper] [code]

• N. Hallak (2026). Sparse mirror descent over symmetric sets: Fixed-point limitations and selected swaps. Submitted to SIAM Journal on Optimization, October 2026. [paper] [code]