ISIT 2023

166 papers
The 2023 edition of the International Symposium on Information Theory (ISIT) featured 179 papers addressing a broad spectrum of information-theoretic research. A substantial number of contributions focused on the intersection of information theory and machine learning, specifically addressing federated learning, algorithmic stability, and generalization. Research directions included the regularization of empirical risk using f-divergences and the analysis of machine learning models beyond the independent and identically distributed (IID) assumption. Technical contributions also emphasized high-dimensional estimation and matrix recovery, with studies on entrywise error analysis for low-rank matrices and spectral filtering for robust estimation. Distributed computing was explored through coded parallelism for deep learning and robust load balancing for heavy-tailed workloads. Other topics included secure gradient aggregation, goal-oriented compression, and error correction for substring edits. Prolific contributors to the year's proceedings included Bo Bai, Antonia Wachter-Zeh, and Deniz Gündüz, reflecting expertise in coding theory, communication systems, and statistical learning.

Top Research Directions

Research directions most represented by ISIT 2023 papers.

Sparsity Preserving Coded Computation 2
Threshold Saturation via Spatially Coupled Codes 2
Algebraic Redundancy for Straggler Resilient Distributed Systems 2
Decoupling Statistical Dependencies via Poissonization 1
Numerical Stability in Coded Distributed Computing 1
Non-Adaptive Experimental Design via Randomized Pooling 1
The Log-Rank Conjecture in Communication Complexity 1
The Computational Overhead of Adaptive Sampling 1
Unsourced Random Access via Shared Codebooks 1
Characterizing Capacity Achieving Input Distributions under Physical Constraints 1
Robust Group Testing for Noisy Measurements 1
Density Evolution for Asymptotic Analysis of Iterative Decoders 1
Proving Streaming Space Lower Bounds via Communication Complexity 1
Information Theoretic Limits of Correlated Graph Alignment 1
Decoding Thresholds and Capacity via Code Symmetry 1

Top Authors

AuthorPapers
Antonia Wachter-Zeh6
Bo Bai6
Deniz Gunduz6
Ilan Shomorony5
Rawad Bitar5
Flavio Calmon4
Maximilian Egger4
Oliver Kosut4
Anuran Makur3
Christina Fragouli3
Cong Shen3
Dor Elimelech3
Olgica Milenkovic3
Shahab Asoodeh3
Aditya Ramamoorthy2
Alejandro Lancho2
Alex Dytso2
Ashish J Khisti2
Banghua Zhu2
Basak Guler2

Impactful Papers

Papers from ISIT 2023 that shaped research clusters.

Influential

Papers

Title
The Privacy-Utility Tradeoff in Rank-Preserving Dataset Obfuscation
The Rate-Distortion-Perception Trade-off with Side Information
The Re-encoding Transform in Algebraic List Decoding of Algebraic Geometric Codes
Thermal-Aware Channel Capacity
Thompson Sampling Regret Bounds for Contextual Bandits with sub-Gaussian rewards
Trading Communication for Computation in Byzantine-Resilient Gradient Coding
Two-Server Private Information Retrieval with Optimized Download Rate and Result Verification
Unconditionally Secure Access Control Encryption
Uniqueness of Distributional BP Fixed Point in Ising Model on Trees
Universal Prediction of m-ary Sequences
Unsourced Massive Access-Based Digital Over-the-Air Computation for Efficient Federated Edge Learning
Variable-Length Insertion-Based Noisy Sorting
Vector Quantization with Error Uniformly Distributed over an Arbitrary Set
Wireless Coded Distributed Learning with Gaussian-based Local Differential Privacy
(αD, αG)-GANs: Addressing GAN Training Instabilities via Dual Objectives
ε-Almost Collision-Flat Universal Hash Functions Motivated by Information-Theoretic Security

« prev Page 4/4