Morteza Zadimoghaddam
I am a staff research scientist (L6) at Google Zurich office in Switzerland. Previously, I spent 4 years in Google Cambridge, two years in the same Zurich office and 4 years in the New York office where I started my career at Google in January 2014. Prior to Google, I did my PhD in computer science at MIT (CSAIL) under supervision of Professor Erik D. Demaine.
I work on applying optimization techniques to various practical problems in order to find provably efficient algorithms. In particular, I apply infrastructure optimization methods to save computational resources at scale. On the mathematical and research side, I am interested in Submodular Optimization and its applications in large scale data mining and machine learning problems.
Authored Publications
Sort By
Preview abstract
In modern datasets, where single records can have multiple owners, enforcing user-level differential privacy requires capping each user's total contribution. This "contribution bounding" becomes a significant combinatorial challenge. Existing sequential algorithms for this task are computationally intensive and do not scale to the massive datasets prevalent today. To address this scalability bottleneck, we propose a novel and efficient distributed algorithm. Our approach models the complex ownership structure as a hypergraph, where users are vertices and records are hyperedges. The algorithm proceeds in rounds, allowing users to propose records in parallel. A record is added to the final dataset only if all its owners unanimously agree, thereby ensuring that no user's predefined contribution limit is violated. This method aims to maximize the size of the resulting dataset for high utility while providing a practical, scalable solution for implementing user-level privacy in large, real-world systems.
View details
GIST: Greedy Independent Set Thresholding for Max-Min Diversification with Submodular Utility
Gui Citovsky
Advances in Neural Information Processing Systems (2025)
Preview abstract
We introduce a novel subset selection problem called \emph{min-distance diversification with monotone submodular utility} (MDMS), which has a wide variety of applications in machine learning, e.g., data sampling and feature selection.
Given a set of points in a metric space, the goal of MDMS is to maximize an objective function combining a monotone submodular utility term and a min-distance diversity term between any pair of selected points, subject to a cardinality constraint.
We propose the $\GIST$ algorithm, which achieves a $\sfrac{1}{2}$-approximation guarantee for MDMS by approximating a series of maximum independent set problems with a bicriteria greedy algorithm.
We also prove that it is NP-hard to approximate to within a factor of $0.5584$.
Finally, we demonstrate that $\GIST$ outperforms existing benchmarks for on a real-world image classification task that studies single-shot subset selection for ImageNet.
View details
Preview abstract
In the differentially private set union problem, users contribute sets of items as input, and the output is a subset of the union of all items. Algorithms for this problem seek to output as many items as possible while maintaining differential privacy with respect to the addition or removal of an individual user.
The basic solution to this problem maintains a weight over each item. Each user contributes uniformly to the items in their set, random noise is added to the weights, and items with noisy weight above a certain threshold are output. The only scalable (i.e., distributed) algorithms for this problem from prior work are this basic algorithm and an iterative method which repeatedly calls the basic algorithm, ignoring items found in prior invocations.
In this work, we give an improved weighting algorithm over basic uniform weighting. Our algorithm reroutes weight from items with weight far above the threshold to items with smaller weight, thereby increasing the probability that less frequent items are output. The algorithm is scalable and does not suffer any privacy loss when compared to the basic algorithm. We prove that our algorithm will never underperform the basic algorithm and show experimentally that replacing the basic algorithm with ours yields the best results among scalable algorithms for the private set union problem.
View details
The Cost of Consistency: Submodular Maximization with Constant Recourse
Paul Duetting
Federico Fusco
Ashkan Norouzi Fard
Ola Svensson
ACM Symposium on Theory of Computing (STOC 2025), ACM, pp. 1406-1417
Preview abstract
In this work, we study online submodular maximization and how the requirement of maintaining a stable solution impacts the approximation. In particular, we seek bounds on the best-possible
approximation ratio that is attainable when the algorithm is allowed to make, at most, a constant number of updates per step. We show a tight information-theoretic bound of $2/3$ for general monotone submodular functions and an improved (also tight) bound of $3/4$ for coverage functions. Since both these bounds are attained by non poly-time algorithms, we also give a poly-time randomized algorithm that achieves a $0.51$-approximation. Combined with an
information-theoretic hardness of $1/2$ for deterministic algorithms from prior work, our work thus shows a separation between deterministic and randomized algorithms, both information theoretically and for poly-time algorithms.
View details
Scalable Private Partition Selection via Adaptive Weighting
Justin Y. Chen
Forty-second International Conference on Machine Learning (2025)
Preview abstract
In the differentially private partition selection problem (a.k.a. private set union, private key discovery), users hold subsets of items from an unbounded universe. The goal is to output as many items as possible from the union of the users' sets while maintaining user-level differential privacy. Solutions to this problem are a core building block for many privacy-preserving ML applications including vocabulary extraction in a private corpus, computing statistics over categorical data and learning embeddings over user-provided items.
We propose an algorithm for this problem, MaxAdaptiveDegree(MAD), which adaptively reroutes weight from items with weight far above the threshold needed for privacy to items with smaller weight, thereby increasing the probability that less frequent items are output. Our algorithm can be efficiently implemented in massively parallel computation systems allowing scalability to very large datasets. We prove that our algorithm stochastically dominates the standard parallel algorithm for this problem. We also develop a two-round version of our algorithm, MAD2R, where results of the computation in the first round are used to bias the weighting in the second round to maximize the number of items output. In experiments, our algorithms provide the best results across the board among parallel algorithms and scale to datasets with hundreds of billions of items, up to three orders of magnitude larger than those analyzed by prior sequential algorithms.
View details
Preview abstract
In the differentially private partition selection problem (a.k.a. private set union, private key discovery), users hold subsets of items from an unbounded universe. The goal is to output as many items as possible from the union of the users' sets while maintaining user-level differential privacy. Solutions to this problem are a core building block for many privacy-preserving ML applications including vocabulary extraction in a private corpus, computing statistics over categorical data and learning embeddings over user-provided items.
We propose an algorithm for this problem, MaxAdaptiveDegree(MAD), which adaptively reroutes weight from items with weight far above the threshold needed for privacy to items with smaller weight, thereby increasing the probability that less frequent items are output. Our algorithm can be efficiently implemented in massively parallel computation systems allowing scalability to very large datasets. We prove that our algorithm stochastically dominates the standard parallel algorithm for this problem. We also develop a two-round version of our algorithm, MAD2R, where results of the computation in the first round are used to bias the weighting in the second round to maximize the number of items output. In experiments, our algorithms provide the best results across the board among parallel algorithms and scale to datasets with hundreds of billions of items, up to three orders of magnitude larger than those analyzed by prior sequential algorithms.
View details
The Cost of Consistency: Submodular Maximization with Constant Recourse
Paul Duetting
Federico Fusco
Ashkan Norouzi Fard
Ola Svensson
Proceedings of the 57th Annual ACM Symposium on Theory of Computing (2025), 1406–1417
Preview abstract
In this work, we study online submodular maximization and how the requirement of maintaining a stable solution impacts the approximation. In particular, we seek bounds on the best-possible
approximation ratio that is attainable when the algorithm is allowed to make, at most, a constant number of updates per step. We show a tight information-theoretic bound of $2/3$ for general monotone submodular functions and an improved (also tight) bound of $3/4$ for coverage functions. Since both these bounds are attained by non poly-time algorithms, we also give a poly-time randomized algorithm that achieves a $0.51$-approximation. Combined with an
information-theoretic hardness of $1/2$ for deterministic algorithms from prior work, our work thus shows a separation between deterministic and randomized algorithms, both information theoretically and for poly-time algorithms.
View details
Deletion Robust Non-Monotone Submodular Maximization over Matroids
Paul Duetting
Federico Fusco
Ashkan Norouzi Fard
Journal of Machine Learning Research, 26 (2025), pp. 1-28
Preview abstract
Maximizing a submodular function is a fundamental task in machine learning and in this paper we study the deletion robust version of the problem under the classic matroids constraint. Here the goal is to extract a small size summary of the dataset that contains a high value independent set even after an adversary deleted some elements. We present constant-factor approximation algorithms, whose space complexity depends on the rank $k$ of the matroid and the number $d$ of deleted elements. In the centralized setting we present a $(4.597+O(\eps))$-approximation algorithm with summary size $O( \frac{k+d}{\eps^2}\log \frac{k}{\eps})$ that is improved to a $(3.582+O(\eps))$-approximation with $O(k + \frac{d}{\eps^2}\log \frac{k}{\eps})$ summary size when the objective is monotone. In the streaming setting we provide a $(9.435 + O(\eps))$-approximation algorithm with summary size and memory $O(k + \frac{d}{\eps^2}\log \frac{k}{\eps})$; the approximation factor is then improved to $(5.582+O(\eps))$ in the monotone case.
View details
Scalable Private Partition Selection via Adaptive Weighting
Justin Y. Chen
Forty-second International Conference on Machine Learning (2025)
Preview abstract
In the differentially private partition selection problem (a.k.a. private set union, private key discovery), users hold subsets of items from an unbounded universe. The goal is to output as many items as possible from the union of the users' sets while maintaining user-level differential privacy. Solutions to this problem are a core building block for many privacy-preserving ML applications including vocabulary extraction in a private corpus, computing statistics over categorical data and learning embeddings over user-provided items.
We propose an algorithm for this problem, MaxAdaptiveDegree(MAD), which adaptively reroutes weight from items with weight far above the threshold needed for privacy to items with smaller weight, thereby increasing the probability that less frequent items are output. Our algorithm can be efficiently implemented in massively parallel computation systems allowing scalability to very large datasets. We prove that our algorithm stochastically dominates the standard parallel algorithm for this problem. We also develop a two-round version of our algorithm, MAD2R, where results of the computation in the first round are used to bias the weighting in the second round to maximize the number of items output. In experiments, our algorithms provide the best results across the board among parallel algorithms and scale to datasets with hundreds of billions of items, up to three orders of magnitude larger than those analyzed by prior sequential algorithms.
View details
Consistent Submodular Maximization
Paul Duetting
Federico Fusco
Ashkan Norouzi Fard
Proceedings of the 41st International Conference on Machine Learning, PMLR (2024), pp. 11979-11991
Preview abstract
Maximizing monotone submodular functions under cardinality constraints is a classic algorithmic problem with several applications in data mining and machine learning. In this paper we study this problem in a dynamic setting with consistency constrains. In this setting, elements arrive in a streaming fashion and one is interested in maintaining a constant approximation to the optimal solution and in having a stable solution (i.e., the number of changes between two consecutive solutions is bounded). We provide several algorithms in this setting with different trade-offs between consistency and approximation quality. We also complement our theoretical results with an experimental analysis showing the effectiveness of our algorithms in real world instances.
View details
