Search
Results
-
Second-Order Asymptotics of Divergence Tests
(2024)International Zurich Seminar on Information and Communication (IZS 2024). ProceedingsConsider a binary statistical hypothesis testing problem, where n independent and identically distributed random variables Z n are either distributed according to the null hypothesis P or the alternate hypothesis Q, and only P is known. A well-known test that is suitable for this case is the so-called Hoeffding test, which accepts P if the Kullback-Leibler (KL) divergence between the empirical distribution of Z n and P is below some ...Conference Paper -
Graph-based Algorithms for Linear Computation Coding
(2024)International Zurich Seminar on Information and Communication (IZS 2024). ProceedingsWe revisit existing linear computation coding (LCC) algorithms, and introduce a new framework that measures the computational cost of computing multidimensional linear functions, not only in terms of the number of additions, but also with respect to their suitability for parallel processing. Utilizing directed acyclic graphs, which correspond to signal flow graphs in hardware, we propose a novel LCC algorithm that controls the tradeoff ...Conference Paper -
Error Probability Trade-off in Quantum Hypothesis Testing via the Nussbaum–Szkoła Mapping
(2024)International Zurich Seminar on Information and Communication (IZS 2024). ProceedingsThe error probability trade-off of quantum hypothesis testing is related to that of a certain surrogate classical hypothesis test via the Nussbaum-Szkoła mapping. This connection was used in the information-theoretic literature to establish the asymptotic error exponent of Bayesian quantum hypothesis testing and asymmetric quantum hypothesis testing (Hoeffding bound). In this work, we analyze the non-asymptotic gap between the error ...Conference Paper -
Universal Neyman–Pearson Classification with a Known Hypothesis
(2024)International Zurich Seminar on Information and Communication (IZS 2024). ProceedingsWe propose a universal classifier for binary Neyman–Pearson classification where null distribution is known while only a training sequence is available for the alternative distribution. The proposed classifier interpolates between Hoeffding’s classifier and the likelihood ratio test and attains the same error probability prefactor as the likelihood ratio test, i.e., the same prefactor as if both distributions were known. Similarly to ...Conference Paper -
A Unified Framework to Generalization Error via Variable-size Compressibility
(2024)International Zurich Seminar on Information and Communication (IZS 2024). ProceedingsIn this paper, we establish novel data-dependent upper bounds on the generalization error through the lens of a “variable-size compressibility” framework that we introduce newly here. In this framework, the generalization error of an algorithm is linked to a variable-size ‘compression rate’ of its input data. This is shown to yield bounds that depend on the empirical measure of the given input data at hand, rather than its unknown ...Conference Paper -
Rate-Loss Regions for Polynomial Regression Side Information
(2024)International Zurich Seminar on Information and Communication (IZS 2024). ProceedingsIn the context of goal-oriented communications, this paper addresses the achievable rate versus generalization error region of a learning task applied on compressed data. The study focuses on the distributed setup where a source is compressed and transmitted through a noiseless channel to a receiver performing polynomial regression, aided by side information available at the decoder. The paper provides the asymptotic rate generalization ...Conference Paper -
Optimizing IRS-Assisted SIMO/MISO Channels: An Analytical Approach
(2024)International Zurich Seminar on Information and Communication (IZS 2024). ProceedingsIntelligent reflective surfaces (IRS) have recently emerged as a tool to improve the energy and spectral efficiencies of wireless systems and networks at reasonable cost. The underlying IRS optimization problems are difficult due to their non-convex nature, complicated analytical structure and the lack of appropriate analytical tools. While a number of algorithms were proposed to obtain locally-optimal or approximate solutions, globally-optimal ...Conference Paper -
M-DAB: An Input-Distribution Optimization Algorithm for Composite DNA Storage by the Multinomial Channel
(2024)International Zurich Seminar on Information and Communication (IZS 2024). ProceedingsRecent experiments have shown that the capacity of DNA storage systems may be significantly increased by synthesizing composite DNA letters. In this work, we model a DNA storage channel with composite inputs as a multinomial channel, and propose an optimization algorithm for its capacity achieving input distribution, for an arbitrary number of output reads. The algorithm is termed multidimensional dynamic assignment Blahut-Arimoto (M-DAB), ...Conference Paper -
Strong Converse for Bi-Static ISAC with Two Detection-Error Exponents
(2024)International Zurich Seminar on Information and Communication (IZS 2024). ProceedingsThe paper considers an information-theoretic bi static integrated sensing and communication (ISAC) model and provides new results on the fundamental limits of this model. Specifically, we show a strong converse result for memoryless ISAC where the channel transition law depends on an underlying binary hypothesis, which the radar receiver wishes to determine with largest exponential decay rates of the probabilities of detection error under ...Conference Paper -
On the Error Exponent Benefit of Sequentiality in Universal Binary Classification
(2024)International Zurich Seminar on Information and Communication (IZS 2024). ProceedingsIn the binary classification problem, the decision maker observes a testing sequence sampled from one of the two unknown distributions P0 and P1, along with two training sequences following each of the distributions. Since the two distributions are unknown, it is natural to impose universal guarantees on certain performances. The focus of this paper is a “semi-sequential” setup where the training sequences have fixed length and testing ...Conference Paper