Graph augmentation plays a crucial role in achieving good generalization for contrastive graph self-supervised learning. However, mainstream Graph Contrastive Learning (GCL) often favors random graph augmentations, by relying on random node dropout or edge perturbation on graphs. Random augmentations may inevitably lead to semantic information corruption during the training, and force the network to mistakenly focus on semantically irrelevant environmental background structures. To address these limitations and to improve generalization, we propose a novel self-supervised learning framework for GCL, which can adaptively screen the semantic-related substructure in graphs by capitalizing on the proposed gradient-based Graph Contrastive Saliency (GCS). The goal is to identify the most semantically discriminative structures of a graph via contrastive learning, such that we can generate semantically meaningful augmentations by leveraging on saliency. Empirical evidence on 16 benchmark datasets demonstrates the exclusive merits of the GCS-based framework. We also provide rigorous theoretical justification for GCS’s robustness properties. Code is available at https://github.com/GCS2023/GCS .

This paper develops a generalized theory of multipath shape factors used to study the small-scale fading statistics in three-dimensional non-omnidirectional multipath channels. Previous results for horizontal propagation channels are special cases of the present framework when the multipath angular power density (APD) and the travel direction of the mobile unit are confined to a plane. The three-dimensional multipath shape factors are derived in terms of the first three degrees of unnormalized spherical harmonics of the APD. Based on the so derived shape factors, analytical expressions for the level crossing rate, average fade duration and approximate envelope correlation are derived for a Rayleigh fading three-dimensional channel.

A novel and general parameterized fading model for the instantaneous received path power is presented, which accounts for both wide-sense stationary shadowing and small-scale fading. A method-of-moments parameter estimator is derived and implemented using a non-linear least-squares algorithm. Performance is explored by analysis and simulation under different noisy channel scenarios, and compared to exact Cramer-Rao bounds in the case of uncorrelated shadowing, and to idealized CR bounds, where shadowing and small-scale fading processes are assumed to be observed separately, in the case when shadowing correlation was not negligible.

Model-based algorithms that attempt to localize targets and estimate the structure within a building using data from external sensors have received much attention in recent years. The potential benefits to homeland security and urban warfare are exceedingly apparent. Accurately estimating the thickness and dielectric constant of the exterior wall could prove to be a critical first step in determining the layout within, i.e., the location of interior walls, doorways, stairwells, etc. However, data collection using a linear sensor arrangement yields an ambiguous two dimensional objective function for the wall parameters, rendering maximum likelihood methods ineffective. We show that a spatially diverse aperture obviates the wall parameter ambiguity and allows accurate estimation of thickness and permittivity using dynamic logic, an iterative model-based approach to maximum likelihood.

In this paper, we investigate the fundamental loss of ranging resolution using received signal strength indicator (RSSI) observations relative to unprocessed received signal (URS) measurements. Unlike previous work, this is done for a single propagation model that accounts for large-scale (log-normal) fading and small-scale (multipath) fading effects. Further, this contribution fully accounts for the temporal correlation of the log-normal attenuation and the multipath channel, which previous work has ignored. Cramer Rao bounds on the estimates of range give us a complete description of the effects of data reduction (through RSSI observations), multipath, and random attenuation on the estimates of range.

Hyperspectral image processing is a pixel-by-pixel approach to the detection and localization of features by spectral analysis techniques. Usually, partial knowledge about the feature, noise, and clutter spectra are provided, and the problem is to 'unmix' each pixel, or to estimate the relative concentrations of the reference spectra on a per pixel basis. A popular method of linear spectral unmixing for hyperspectral imagery is linear least squares. Linear least square approaches are appropriate when observational errors predominate and are inappropriate when significant modeling errors are present. The least square approach has some disadvantages, especially in cases with few, poorly known references or significant reference variation throughout an image. In this article, the restricted total least squares (RTLS) approach is presented and evaluated on experimental data. Although the proposed RTLS require more calculations than linear least squares, its relative error performance is much better.

The asymptotic multiuser efficiencies (AMEs) are derived for various classes of decision-directed multiuser detectors, including multistage detectors, and decision-feedback detectors. Novel classes of soft-decision multistage detectors are proposed and analyzed. Each class is specified in part by a soft-decision nonlinearity, such as a symmetric quantizer or a linear clipper. Closed-form expressions for two-user AMEs are derived for soft-decision two-stage detectors and can be used as a design criterion to optimize the soft-decision nonlinearities. For a special case of two synchronous users, the soft-decision two-stage detector using an optimized linear clipper with either conventional or decorrelated tentative decisions is shown to achieve optimum AME. Upper and lower bounds on the AME are obtained for decision-feedback detectors using either conventional or decorrelated tentative decisions. It is demonstrated that decision-directed multiuser detectors with conventional tentative decisions have low near-far resistance compared to those with decorrelated tentative decisions.

Analytical expressions for the error probability of linear multipath-decorrelating receivers with coherent and differentially coherent reception are derived. Both multi-user receivers have superior performance compared to the RAKE receiver and eliminate error probability floor caused by multiple-access interference on a code-division multiple access (CDMA) reverse link. Results also emphasize benefits of using coherent multi-user reception with maximal-ratio combining in frequency-selective fading channels.

We consider an additive white Gaussian noise multiple-access channel and investigate the performance degradation of several multiuser detectors due to imperfect knowledge of the channel parameters, induced by either estimation errors or sampling/quantization of the receiver's sufficient statistics. The symbol error rates and the asymptotic multiuser efficiencies (AMEs) of these detectors are examined as functions of both true and approximated channel parameters for 2 asynchronous users. The results quantify the loss of near-far resistance due to parameter estimation errors and demonstrate that detectors which are optimal for ideal conditions are often more sensitive to parameter mismatch. Examples are given which illustrate a reverse ordering of the symbol error rates for these detectors in the high signal-to-noise regime, and also illustrate cases when the conventional matched filter detector exhibits superior performance over multiuser detectors.

Proposes a suboptimal low-complexity multiuser receiver for synchronous CDMA frequency-selective Rayleigh fading channels. In contrast to the conventional RAKE receiver, which suffers from near-far effects due to channel fading, the proposed multiuser receiver is shown to alleviate the near-far problem while preserving multipath diversity gain. This is demonstrated by comparing the symbol error probability and asymptotic multiuser efficiency of the proposed multiuser detector and RAKE receiver.>

Presents a performance analysis of a decorrelating detector for differentially coherent transmission over a CDMA frequency-nonselective Rayleigh fading channel. While the correlation-type receiver is near-far limited in this channel, it is shown that the aforementioned multiuser detector eliminates the bit error rate floor and alleviates the near-far problem.>

The acoustic local area network (ALAN) is designed for multipoint-to-point telemetry between ocean bottom sensors and a surface receiver in deep water areas. The main obstacle for coherent communications in this channel is strong multiple-access interference from cochannel signals, which affects a common narrowband request channel. Throughput and packet delay in this network depends on the capability of the receiver to resolve collisions between request packets. In this paper we propose an adaptive multiuser receiver for joint parameter estimation and multiple-access interference cancellation. It extends previous work on adaptive multiuser detectors to the more realistic communication scenario by estimating signal parameters rather than assuming that they are known. The adaptive approach is different from the one used in a centralized multiuser receiver in that no knowledge of the signature sequences of any of the active users is required.>

Analyzes the throughput of a centralized packet receiver which utilizes multiuser detection in an unslotted contention channel. The receiver is capable of demodulating all overlapping, unit-length packets, provided that the T-length packet headers do not overlap, and at most K packets overlap at any one time. If one of these conditions is violated, the remainder of the busy period is assumed to be unresolvable. Assuming that the aggregate packet arrival process is Poisson, the authors analyze the throughput as a function of the total offered load (/spl lambda/), the maximum number of resolvable overlapping packets (K), and the fractional header duration (T). They derive an iterative expression for the throughput for K=2 or 3, generalizable to larger K, which permits solution of the throughput to arbitrary accuracy. A closed form solution is given for the case K=2 and T=0.5. The authors provide bounds on the throughput for arbitrary K, /spl lambda/, and T, which are more convenient than the exact approach for T>0 or large K. Numerical results illustrate the improvement in throughput due to multiuser detection, and suggest a technique for finding the minimum complexity multiuser receiver (minimum K) as a function of other system parameters.>

The performance analysis of maximum likelihood sequence and decorrelating multiuser detectors is developed for the single-path Rayleigh fading channel. The received signal in this multipoint-to-point channel is additively composed of white Gaussian noise and independently faded waveforms from a finite number of asynchronous users. The optimum sequence detector consists of a bank of single-user matched filters followed by a dynamic programming algorithm, has a demodulation complexity per symbol which is exponential in the number of users, and requires the knowledge of the system as well of the fading parameters. The decorrelating detector has a complexity which is polynomial in the number of users, and does not require knowledge of the signal amplitudes, which are assumed to vary slowly relative to the symbol rate. The single-user error rates are presented for both detectors and compared to conventional detection and isolated transmission, and the notion of asymptotic multiuser efficiency is developed for this channel. In contrast to conventional detection; which suffers from near-far effects due to Rayleigh fading, it is shown that the aforementioned multiuser detectors exhibit performance which rivals that of isolated transmission for sufficiently small thermal noise power. The results of the analysis are illustrated with numerical examples and indicate that the near-far problem may be alleviated by a more sophisticated receiver design.>

The performance analysis of low-complexity decorrelating detectors is presented for both coherent and differentially coherent transmission over the multiple-access, frequency-nonselective Rayleigh fading channel. While it is known that correlation-type receivers are near-far limited in the case when a single fading path dominates, it is shown that the multiuser detectors eliminate the bit error rate floor, alleviate the near-far problem, and rival optimum demodulation for isolated transmission. The results apply to wideband and narrowband CDMA systems.>

A model noncoherent, optical, asynchronous, CDMA system is described. The error rate for a single-user matched-filter receiver that is valid for arbitrary photomultipliers and signature sequence sets, adheres to the semiclassical model of light, and does not depend on approximations for large user groups, strong received optical fields, or chip synchronism is analyzed. The exact minimum probability of error and optimal threshold are compared to those obtained with user-synchronism and multiple-access interference (MAI) distribution approximations. For the special case of unity-gain photodetectors and prime sequences, it is shown that the approximation of chip synchronism yields a weak upper bound on the exact error rate. It is demonstrated that the approximations of perfect optical-to-electrical conversion and Gaussian-distributed MAI yield a poor approximation to the minimum error rate and an underestimate of the optimal threshold. Arbitrarily tight bounds are developed on the error rate for unequal energies per bit. In the case when the signal energies coincide, these bounding expressions are considerably easier to compute than the exact error rate.>

An optical, direct-detection differential phase-shift keying (DPSK) receiver whose error probability is quantum-limited as the transmitting laser linewidth vanishes is analyzed. The receiver design is based on a binary equiprobable hypothesis test with doubly stochastic point process observations, the conditional random rates of which depend on the transmitting laser phase noise, which is modeled as a Brownian motion. The receiver structure consists of a simple delay-and-sum optical preprocessor followed by a photoelectric converter and an integrate-and-dump circuit. Upper and lower bounds on the receiver bit error rate are derived by developing bounds on the conditional rates of the point process, and it is shown that the error probability bounds converge to the true value as the transmitting laser linewidth decreases. Bounds on the power penalty are computed for parameters corresponding to existing semiconductor injection lasers, and are seen to be less than the limiting power penalty for the balanced DPSK receiver.>