We develop elements of a theory cooperation and coordination in networks. Rather than considering communication network as means distributing information, or reconstructing random processes at remote nodes, we ask what dependence can be established among the nodes given constraints. Specifically, with rates {R_{i,j}} between is set all achievable joint distributions p(x1, ..., xm) actions network. Several networks are solved, including arbitrarily large cascade Distributed solution to many...

We consider capacity of discrete-time channels with feedback for the general case where is a time-invariant deterministic function output samples. Under assumption that channel states take values in finite alphabet, we find sequence achievable rates and upper bounds on capacity. The are computable any N, limits sequences exist. show when probability initial state positive all states, then limit achievable-rate sequence. further stationary, indecomposable, has no intersymbol interference...

We establish that the feedback capacity of trapdoor channel is logarithm golden ratio and provide a simple communication scheme achieves capacity. As part analysis, we formulate class dynamic programs characterize capacities unifilar finite-state channels. The an instance admits closed-form solution.

We investigate the role of directed information in portfolio theory, data compression, and statistics with causality constraints. In particular, we show that is an upper bound on increment growth rates optimal portfolios a stock market due to causal side information. This tight for gambling horse race, which extreme case markets. Directed also characterizes value instantaneous compression quantifies benefit inference joint two stochastic processes. hypothesis testing, evaluates best error...

Four estimators of the directed information rate between a pair jointly stationary ergodic finite-alphabet processes are proposed, based on universal probability assignments. The first one is Shannon--McMillan--Breiman type estimator, similar to those used by Verd\'u (2005) and Cai, Kulkarni, (2006) for estimation other measures. We show almost sure $L_1$ convergence properties estimator any underlying assignment. three map assignments different functionals, each exhibiting relative merits...

The secrecy capacity of the type II wiretap channel (WTC II) with a noisy main is currently an open problem. Herein its secrecy-capacity derived and shown to be equal semantic-security (SS) capacity. In this setting, legitimate users communicate via discrete-memoryless (DM) in presence eavesdropper that has perfect access subset choosing transmitted symbols, constrained fixed fraction blocklength. criterion achieved simultaneously for all possible choices. SS demands negligible mutual...

We introduce Gaussian mixture models of 'structure' and colour features in order to classify coloured textures images, with a view the retrieval textured images from databases. Classifications are performed separately using structure then combined confidence criterion. apply VisTex database classification man-made natural areas aerial images. compare these others literature, show an overall improvement performance.

We study source coding in the presence of side information, when system can take actions that affect availability, quality, or nature information. begin by extending Wyner-Ziv problem with decoder information to case where is allowed choose affecting then consider setting are taken encoder, based on its observation source. Actions may have costs commensurate quality they yield, and an overall per-symbol cost constraint be imposed. characterize achievable tradeoffs between rate, distortion,...

The Ising channel, which was introduced in 1990, is a channel with memory that models intersymbol interference. In this paper, we consider the feedback and find capacity of together capacity-achieving coding scheme. To calculate capacity, an equivalent dynamic programming (DP) problem formulated solved. Using DP solution, establish expression C = (2H <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">b</sub> (a)/3+a) ≈ 0.575522, where (a)...

A notion of directed information between two continuous-time processes is proposed. key component in the definition taking an infimum over all possible partitions time interval, which plays a role no less significant than supremum "space" inherent mutual information. Properties and operational interpretations estimation communication are then established for proposed For additive white Gaussian noise channel, it shown that Duncan's classical relationship causal error continues to hold...

Determining a single-letter secrecy-capacity formula for the arbitrarily varying wiretap channel (AVWTC) is an open problem largely because of two main challenges. Not only does it capture difficulty compound (another problem), also requires that secrecy ensured with respect to exponentially many possible state sequences. By extending strong soft-covering lemma, recently derived by authors, heterogeneous scenario, this paper accounts exponential number constraints while imposing on...

We consider finite state channels, where the of channel is its previous output. refer to these as Previous Output STate (POST) channels. first focus on POST(α) These channels have binary inputs and outputs, determines if behaves a Z or an S channel, both with parameter α. show that nonfeedback capacity equals feedback capacity, despite memory channel. The proof this surprising result based showing induced output distribution, when maximizing directed information in presence feedback, can...

This work addresses the problem of direction-of-arrival (DOA) estimation in presence non-Gaussian, heavy-tailed, and spatially-colored interference. Conventionally, interference is considered to be Gaussian-distributed spatially white. However, practice, this assumption not guaranteed, which results degraded DOA performance. Maximum likelihood non-Gaussian computationally complex practical. Therefore, proposes a neural network (NN)-based approach for spatial spectrum multisource scenarios...

This work develops a new method for estimating and optimizing the directed information rate between two jointly stationary ergodic stochastic processes. Building upon recent advances in machine learning, we propose recurrent neural network (RNN)-based estimator which is optimized via gradient ascent over RNN parameters. The does not require prior knowledge of underlying joint/marginal distributions can be easily continuous input processes realized by deep generative model. We prove...

The detection of deepfake speech has become increasingly challenging with the rapid evolution technologies. In this paper, we propose a hybrid architecture for detection, combining self-supervised learning framework feature extraction classifier head to form an end-to-end model. Our approach incorporates both audio-level and feature-level augmentation techniques. Specifically, introduce analyze various masking strategies augmenting raw audio spectrograms enhancing representations during...

The capacity of a channel with an energy-harvesting (EH) encoder and finite battery remains open problem, even in the noiseless case. A key instance this scenario is binary EH (BEHC), where has unit-sized inputs. Existing expressions for BEHC are not computable, motivating work, which determines to any desired precision via convex optimization. By modeling system as finite-state state information known causally at encoder, we derive single-letter lower upper bounds using auxiliary directed...

Despite the widespread use of information measures in analyzing probabilistic systems, effective visualization tools for understanding complex dependencies sequential data are scarce. In this work, we introduce matrix (InfoMat), a novel and intuitive representation transfer systems. InfoMat provides structured visual perspective on mutual decompositions, enabling discovery new relationships between enhancing interpretability time series analytics. We demonstrate how captures key measures,...

<para xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> The capacity region of the finite-state multiple-access channel (FS-MAC) with feedback that may be an arbitrary time-invariant function output samples is considered. We characterize both inner and outer bound for this region, using Massey's directed information. These bounds are shown to coincide, hence yield indecomposable FS-MACs without stationary feedback, where state process not affected by...

We investigate the capacity of a multiple access channel with cooperating encoders where partial state information is known to each encoder and full decoder. The cooperation between has two-fold purpose: generate empirical coordination encoders, share about private messages that has. For two-way cooperation, this purpose achieved by double-binning, first layer binning used similarly source coding, second transmit messages. complete result provides framework perspective for addressing complex...

The input-constrained erasure channel with feedback is considered, where the binary input sequence contains no consecutive ones, i.e., it satisfies (1, ∞)-RLL constraint. We derive capacity for this setting, which can be expressed as C <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ε</sub> = max xmlns:xlink="http://www.w3.org/1999/xlink">0≤ p≤0.5</sub> ((1-ε)H xmlns:xlink="http://www.w3.org/1999/xlink">b</sub> (p))/(1+(1-ε)p) , ε probability...

An enhanced framework for peak-to-average power ratio (PAPR) reduction and waveform design Multiple-Input-Multiple-Output (MIMO) orthogonal frequency-division multiplexing (OFDM) systems, based on a convolutional-autoencoder (CAE) architecture, is presented. The end-to-end learning-based autoencoder (AE) communication networks represents the network by an encoder decoder, where in between, learned latent representation goes through physical channel. We introduce joint learning scheme...