Search

In this paper, we analyze the generalization performance of the Iterative Hard Thresholding (IHT) algorithm widely used for sparse recovery problems. The parameter estimation and sparsity recovery consistency of IHT has long been known in compressed sensing. From the perspective of statistical learning, another fundamental question is how well the IHT estimation would predict on unseen data. This paper makes progress towards answering this open question by introducing a novel sparse generalization theory for IHT under the notion of algorithmic stability. Our theory reveals that: 1) under natural conditions on the empirical risk function over $n$ samples of dimension $p$ , IHT with sparsity level $k$ enjoys an $\tilde {\mathcal {O}}(n^{-1/2}\sqrt {k\log (n)\log (p)})$ rate of convergence in sparse excess risk; 2) a tighter $\tilde {\mathcal {O}}(n^{-1/2}\sqrt {\log (n)})$ bound can be established by imposing an additional iteration stability condition on a hypothetical IHT procedure invoked to the population risk; and 3) a fast rate of order $\tilde {\mathcal {O}}\left ({n^{-1}k(\log ^{3}(n)+\log (p))}\right)$ can be derived for strongly convex risk function under proper strong-signal conditions. The results have been substantialized to sparse linear regression and sparse logistic regression models to demonstrate the applicability of our theory. Preliminary numerical evidence is provided to support our theoretical predictions. [ABSTRACT FROM AUTHOR]

Contributions: The underrepresentation of women in computer science (CS) majors has long been a focus of attention in many academic documents, the majority of them from the United States and Europe. There is, however, a lack of information about educational interventions (EIs) for women in computing in Latin America. The contribution of this article is to cover this gap and describe what researchers in Latin American countries have been publishing about the recruitment and retention of women in the CS field. Background: Many EIs targeting female students at different educational levels—K-12, undergraduate, and graduate—have been undertaken to increase the participation of women in computing in Latin America. However, descriptions of these activities rarely are included in international academic databases. Research Questions: This literature mapping addresses two main research questions (RQ) about the topic of women in computing in academic publications in Latin American countries: RQ1) what are the characteristics of the publications about women in computing in Latin America? and RQ2) what are the published interventions to recruit and retain women in computing in Latin America? To answer RQ1, six subquestions were created covering year, language, country of origin, document type, and professional track. Furthermore, for RQ2, two subquestions were created involving educational level and the use of software development with a female theme. Methodology: This investigation used the systematic literature mapping process. To achieve a broad coverage of papers, the following sources were included: Scopus, Web of Science, Google Scholar, EBSCO, the proceedings of the Latin American Women in Computing Conference (LAWCC), and those of the Women in Technology (WIT) workshop colocated with the annual conference of the Brazilian Computer Society (SBC). The included papers were published in the last decade (2010–2020) and written in English, Portuguese, or Spanish. Findings: The literature mapping encompasses 197 academic documents, 48.2% of which were written in Portuguese, 28.7% in English, and 23.1% in Spanish. The papers originated from 15 of the 20 Latin American countries. Brazil and Costa Rica have the highest number of publications overall. The documents describe initiatives to increase the participation of women in computing majors that cover the entire educational spectrum, from K-12 to graduate programs, but papers targeting populations in higher education have started to appear recently. [ABSTRACT FROM AUTHOR]

This paper explores the feasibility and scope of co-channel coexistence between a low data rate Orthogonal Frequency Division Multiplexing (OFDM) secondary network and a digital television (DTV) broadcast network operating in the same area. Since DTV receivers do not transmit, neither the presence nor the impact of interference from the secondary network on these DTV receivers can be known. Consequently, conventional spectrum sensing techniques cannot be employed effectively. This paper proposes to compensate the lack of feedback by combining the available information about population distribution in the region - ranging from coarse population density estimates to accurate geospatial maps - with better estimation of interference using a data-driven technique called Ordinary Kriging. Radio Environment Maps constructed using the Kriging method from field-reported observations from the secondary nodes are used to obtain better accuracy than the conventional statistical models. The estimates are then used to compute communication parameters for each secondary link in terms of power, number of active subcarriers, and active time slots, while keeping the average interference in the DTV network within threshold bounds. Thus, the viability of such co-channel license-exempt operation of secondary networks in the DTV black and grey spaces for low data rate applications in densely populated urban areas and higher data rate capabilities for sparsely populated rural areas is established. [ABSTRACT FROM AUTHOR]

This paper considers the problem of adaptive group testing for isolating up to $k$ defective items from a population of size $n$. There exist restrictions or preferences which determine how the items can be pooled for testing. A graphical model formalizes the pooling restrictions and preferences. Such graph-constrained group testing is investigated in three settings: populations with defectives, populations facing the potential presence of inhibitors, and populations with community structures. Adaptive group testing frameworks are provided for each setting. In populations without inhibitors, existing non adaptive frameworks can isolate the defective items perfectly with $\Theta(k \log n/k)$ number of tests, where is the $\beta$ -mixing time of a random walk over the underlying graph. This paper provides a two-stage framework that can perfectly isolate up to $k$ defective items for a regular graph using $\Theta(k2 \log n k + k)$ number of tests, thus achieving an approximate gain of a factor of $k$ over the non-adaptive frameworks. This twostage framework's principles are extended to community-structured graphs and graphs with up to $r$ inhibitor items. In particular, when inhibitors are present in the graph, a four-stage group testing framework is proposed. The results show that in the regime $r= O(k)$ for a fully connected graph, $\Theta ((k+r)\log n/(k+r) + r\log n)$ tests are sufficient for isolating the defective items. This matches the corresponding necessary condition on tests which scales $(k+r)\log n$. The adaptive graphconstrained group testing framework is also empirically evaluated. [ABSTRACT FROM AUTHOR]

Let $\mathrm {W}\in \mathbb {C}^{n\times n}$ be a single-spiked Wishart matrix in the class $\mathrm {W}\sim \mathcal {CW}_{n}(m,\mathrm {I}_{n}+ \theta \mathrm {v}\mathrm {v}^{\dagger}) $ with $m\geq n$ , where ${\mathrm {I}}_{n}$ is the $n\times n$ identity matrix, $\mathrm {v}\in \mathbb {C}^{n\times 1}$ is an arbitrary vector with unit Euclidean norm, $\theta \geq 0$ is a non-random parameter, and $(\cdot)^{\dagger} $ represents the conjugate-transpose operator. Let u1 and ${\mathrm {u}}_{n}$ denote the eigenvectors corresponding to the smallest and the largest eigenvalues of W, respectively. This paper investigates the probability density function (p.d.f.) of the random quantity $Z_{\ell }^{(n)}=\left |{\mathrm {v}^{\dagger} \mathrm {u}_\ell }\right |^{2}\in (0,1)$ for $\ell =1,n$. In particular, we derive a finite dimensional closed-form p.d.f. for $Z_{1}^{(n)}$ which is amenable to asymptotic analysis as $m,n$ diverges with $m-n$ fixed. It turns out that, in this asymptotic regime, the scaled random variable $nZ_{1}^{(n)}$ converges in distribution to $\chi ^{2}_{2}/2(1+\theta)$ , where $\chi _{2}^{2}$ denotes a chi-squared random variable with two degrees of freedom. This reveals that u1 can be used to infer information about the spike. On the other hand, the finite dimensional p.d.f. of $Z_{n}^{(n)}$ is expressed as a double integral in which the integrand contains a determinant of a square matrix of dimension $(n-2)$. Although a simple solution to this double integral seems intractable, for special configurations of $n=2,3$ , and 4, we obtain closed-form expressions. [ABSTRACT FROM AUTHOR]

Understanding intra-urban human mobility patterns and their potential driving forces are vital to city planning and commercial site selection. In this paper, we first investigate the functions of urban regions and how different region types dynamically influence people’s trip decisions. Furthermore, we characterize urban circadian rhythms by time-vary inter-regional transition probabilities between these regions with different functions, and integrate them into intervening opportunity model to predict human mobility. Public transportation card data in Shanghai are used to demonstrate the effectiveness of the model in terms of station passenger flows, travel time and trip flux. By taking regional function into consideration, the proposed model significantly improved the prediction accuracy. Quantitative analysis ulteriorly indicates that trip intentions and regional features are critical elements in trip flux prediction, especially in the afternoon and evening when people have an abundance of opportunities to travel by their own volition. When the function of a certain region changes, our model is able to make reasonable predictions accordingly. The results indicate the importance of considering individual travel motivation and regional function in modeling human mobility. The proposed model could serve as a guide for popularity and trip flux prediction in urban planning and reconstruction. [ABSTRACT FROM AUTHOR]

Pattern mining (PM) refers to the process of discovering patterns of interest to users from data. However, most studies have considered only one pattern, such as frequent pattern or high-utility pattern. With the continuous requirement of businesses in various industries, the single-objective PM methods are difficult to meet the increasingly diverse needs of users. In this paper, a multi-objective problem model for high quality pattern mining (HQPM) is proposed, where the objectives are support, occupancy, and utility. In order to solve the proposed three-objective problem efficiently, an improved multi-objective evolutionary algorithm for HQPM (MOEA-PM) is proposed. Two kinds of population initialization strategies are designed, which is used to ensure the population is effectively distributed in the feasible solution space. By taking the properties of the model into consideration, an auxiliary tool is proposed to accelerate the convergence of the algorithm. Experimental results on real-world datasets show that the proposed three-objective problem model with the MOEA-PM algorithm can discover patterns that are both frequently occurring and has a high utility in the transaction datasets, while at the same time being relatively complete. Compared with the state-of-the-art MOEA-based HQPM algorithms, MOEA-PM has better performance in terms of efficiency, quality, and convergence speed. [ABSTRACT FROM AUTHOR]

Knowledge graph is becoming an indispensable resource that offers structured information for numerous AI applications. However, the knowledge graph often suffers from its incompleteness. Building a complete, high-quality knowledge graph is time-consuming and requires significant human annotation efforts. In this paper, we study the Knowledge Graph Population task, which aims at extending the scale of structured knowledge, with a special focus on reducing data preparation and annotation efforts. Previous works mainly based on discriminative methods build classifiers and verify candidate triplets that are extracted from texts, which heavily rely on the quality of data collection and co-occurrance of entities in the text. However, such methods fail to generalize on entity pairs that are not highly co-occurred, and fail to discover entity pairs that are not co-occurred at all in the given text corpus. We introduce a generative perspective to approach this task and define each relationship by learning the data distribution that embodies the core common properties for relational reasoning. A generative model KGGen is proposed, which samples from the learned data distribution for each relation and can generate triplets regardless of entity pair co-occurrence in the text corpus. To further improve the generation quality while alleviate human annotation efforts, adversarial learning is adopted to not only encourage generating high quality triplets, but also give model the ability to automatically assess the generation quality. Quantitative and qualitative experimental results conducted on two real-world generic knowledge graphs show that the proposed model KGGen generates novel and meaningful triplets with improved efficiency and less human annotation comparing with the state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

Vehicle-to-vehicle (V2V) communications is an important technology in vehicular ad hoc network (VANET) to support autonomous data exchange among vehicles. Multiple V2V communications modes have been investigated for VANET, including dedicated short-range communications (DSRC) based on IEEE 802.11p and LTE-V2X, which are suitable for different packet transmission cases. In order to fully exploit the strengths of various modes, hybrid V2V communications strategies are designed in this paper, where each vehicle is allowed to choose different modes for different kinds of transmissions in separate frequency bands for transmission throughput improvement. Furthermore, since it is usually impractical to decide all the vehicles’ communications strategies globally due to high computational complexity and heavy overhead to broadcast the decisions, we model the selection of hybrid V2V communications strategies for vehicles as an evolutionary game. A strategy selection algorithm is then proposed, where each vehicle can select its hybrid V2V communications strategy locally based on its evaluation of the payoffs for different strategies and limited signalling from the base station. Simulation results demonstrate that the proposed algorithm can converge to an asymptotic stable state, which can improve the transmission throughput of vehicles, and is robust to the slight evaluation errors of payoffs and the strategy mutations of a few vehicles. [ABSTRACT FROM AUTHOR]

With the ever-increasing urbanization process, modeling people's spatiotemporal activities from their online traces has become a crucial task. State-of-the-art methods for this task rely on cross-modal embedding, which maps items from different modalities (e.g., location, time, text) into the same latent space. Despite their inspiring results, existing cross-modal embedding methods merely capture co-occurrences between items without modeling their high-order interactions. In this paper, we first construct two graphs from raw data records to represent the user interaction graph layer and activity graph layer and propose a hierarchical cross-modal embedding method that takes the high-order relationships into consideration. The key notion behind our method is a novel hierarchical embedding framework with meta-graphs connecting different layers. We introduce both inter-record and intra-record meta-graph structures, which enable learning distributed representations that preserve high-order proximities across graphs from different layers. Our empirical experiments on three real-world datasets demonstrate that our method not only outperforms state-of-the-art methods for spatiotemporal activity prediction, but also captures cross-modal proximity at a finer granularity. [ABSTRACT FROM AUTHOR]

The $k\text {-means}$ clustering problem concerns finding a partition of the data points into $k$ clusters such that the total within-cluster squared distance is minimized. This optimization objective is non-convex, and not everywhere differentiable. In general, there exist spurious local solutions other than the global optimum. Moreover, the simplest and most popular algorithm for $k\text {-means}$ , namely Lloyd’s algorithm, generally converges to such spurious local solutions both in theory and in practice. In this paper, we investigate the structures of these spurious local solutions under a probabilistic generative model with $k$ ground truth clusters. As soon as $k=3$ , spurious local minima provably exist, even for well-separated clusters. One such local minimum puts two centers at one true cluster, and the third center in the middle of the other two true clusters. We prove that this is essentially the only type of spurious local minima under a separation condition. In particular, any local minimum solution only involves a configuration that puts multiple centers at a true cluster, and one center in the middle of multiple true clusters. Our results pertain to the $k\text {-means}$ formulation for mixtures of Gaussians or bounded distributions, and hold in the over- and under-parametrization regimes where the number of centers in $k\text {-means}$ may not equal to the number of true clusters. Our theoretical results corroborate existing empirical observations and provide justification for popular heuristics for $k\text {-means}$ clustering. [ABSTRACT FROM AUTHOR]

Stochastic gradient descent (SGD) has become the method of choice for training highly complex and nonconvex models since it can not only recover good solutions to minimize training errors but also generalize well. Computational and statistical properties are separately studied to understand the behavior of SGD in the literature. However, there is a lacking study to jointly consider the computational and statistical properties in a nonconvex learning setting. In this paper, we develop novel learning rates of SGD for nonconvex learning by presenting high-probability bounds for both computational and statistical errors. We show that the complexity of SGD iterates grows in a controllable manner with respect to the iteration number, which sheds insights on how an implicit regularization can be achieved by tuning the number of passes to balance the computational and statistical errors. As a byproduct, we also slightly refine the existing studies on the uniform convergence of gradients by showing its connection to Rademacher chaos complexities. [ABSTRACT FROM AUTHOR]

Group testing refers to the process of testing pooled samples to reduce the total number of tests. Given the current pandemic, and the shortage of test supplies for COVID-19, group testing can play a critical role in time and cost efficient diagnostics. In many scenarios, samples collected from users are also accompanied with auxiliary information (such as demographics, history of exposure, onset of symptoms). Such auxiliary information may differ across patients, and is typically not considered while designing group testing algorithms. In this paper, we abstract such heterogeneity using a model where the population can be categorized into clusters with different prevalence rates. The main result of this work is to show that exploiting knowledge heterogeneity can further improve the efficiency of group testing. Motivated by the practical constraints and diagnostic considerations, we focus on two-stage group testing algorithms, where in the first stage, the goal is to detect as many negative samples by pooling, whereas the second stage involves individual testing to detect any remaining samples. For this class of algorithms, we prove that the gain in efficiency is related to the concavity of the number of tests as a function of the prevalence. We also show how one can choose the optimal pooling parameters for one of the algorithms in this class, namely, doubly constant pooling. We present lower bounds on the average number of tests as a function of the population heterogeneity profile, and also provide numerical results and comparisons. [ABSTRACT FROM AUTHOR]

The basic goal of threshold group testing is to identify up to $d$ defective items among a population of $n$ items, where $d$ is usually much smaller than $n$. The outcome of a test on a subset of items is positive if the subset has at least $u$ defective items, negative if it has up to $\ell $ defective items, where $0 \leq \ell < u$ , and arbitrary otherwise. This is called threshold group testing. The parameter $g = u - \ell - 1$ is called the gap. In this paper, we focus on the case $g > 0$ , i.e., threshold group testing with a gap. Note that the results presented here are also applicable to the case $g = 0$ ; however, the results are not as efficient as those in related work. Currently, a few reported studies have investigated test designs and decoding algorithms for identifying defective items. Most of the previous studies have not been feasible because there are numerous constraints on their problem settings or the decoding complexities of their proposed schemes are relatively large. Therefore, it is compulsory to reduce the number of tests as well as the decoding complexity, i.e., the time for identifying the defective items, for achieving practical schemes. The work presented here makes five contributions. The first is a more accurate theorem for a non-adaptive algorithm for threshold group testing proposed by Chen and Fu. The second is an improvement in the construction of disjunct matrices, which are the main tools for tackling (threshold) group testing and other tasks such as constructing cover-free families or learning hidden graphs. Specifically, we present a better exact upper bound on the number of tests for disjunct matrices compared with that in related work. The third and fourth contributions are a reduced exact upper bound on the number of tests and a reduced asymptotic bound on the decoding time for identifying defective items in a noisy setting on test outcomes. The fifth contribution is a simulation on the number of tests of the resulting improvements for previous work and the proposed theorems. [ABSTRACT FROM AUTHOR]

This paper studies the problem of developing an efficient signal sampling framework for distributed blind spectrum sensing, together with the corresponding local cognitive radio (CR) decision and fusion rules. A novel collaborative sampling and binary output (CSBO) generating scheme is proposed to overcome the disadvantage of traditional energy detectors for blind spectrum sensing, in which the signal-to-noise ratio (SNR) and level of noise power at the CRs are unknown and thus the decision threshold of local energy detectors cannot be specified. In designing the fusion rule in conjunction with the proposed CSBO scheme, both fixed-sample-size tests and sequential tests are considered. It is theoretically shown that the proposed CSBO scheme outperforms traditional blind cooperative energy detectors with uniform sampling schemes in terms of its ability to increase the amount of information acquired under a given number of samples. Simulation results confirm the theoretical analyses, and demonstrate that, compared with the fixed-sample-size test, the sequential test requires fewer observations on average to achieve comparable detection performance. The reduction in the expected sample size needed in the sequential CSBO test when compared to the fixed-sample-size CSBO can reach 50% when primary signal is present. [ABSTRACT FROM AUTHOR]

Power cables are critical assets for the reliable operation of the grid. The cable lifetime is generally estimated from the conductor temperature, and associated lifetime reduction. However, these tasks are intricate due to the complex physics-of-failure (PoF) degradation mechanism of the cable. This is further complicated with the different sources of uncertainty that affect the cable lifetime estimation. Generally, simplified or deterministic PoF models are adopted resulting in non-accurate decision-making under uncertainty. In contrast, the integration of uncertainties leads to a probabilistic decision-making process impacting directly on the flexibility to adopt decisions. Accordingly, this paper presents a novel cable lifetime estimation framework that connects data-driven probabilistic uncertainty models with PoF-based operation, and degradation models through Bayesian state-estimation techniques. The framework estimates the cable health state, and infers confidence intervals to aid decision-making under uncertainty. The proposed approach is validated with a case study with different configuration parameters, and the effect of measurement errors on cable lifetime are evaluated with a sensitivity analysis. Results demonstrate that ambient temperature measurement errors influence more than load measurement errors, and the greater the cable conductor temperature the greater the influence of uncertainties on the lifetime estimate. [ABSTRACT FROM AUTHOR]

In many diagnostic imaging settings, including positron emission tomography (PET), images are typically used for multiple tasks such as detecting disease and quantifying disease. Unlike conventional image reconstruction that optimizes a single objective, this work proposes a multi-objective optimization algorithm for PET image reconstruction to identify a set of images that are optimal for more than one task. This work is reliant on a genetic algorithm to evolve a set of solutions that satisfies two distinct objectives. In this paper, we defined the objectives as the commonly used Poisson log-likelihood function, typically reflective of quantitative accuracy, and a variant of the generalized scan-statistic model, to reflect detection performance. The genetic algorithm uses new mutation and crossover operations at each iteration. After each iteration, the child population is selected with non-dominated sorting to identify the set of solutions along the dominant front or fronts. After multiple iterations, these fronts approach a single non-dominated optimal front, defined as the set of PET images for which none the objective function values can be improved without reducing the opposing objective function. This method was applied to simulated 2D PET data of the heart and liver with hot features. We compared this approach to conventional, single-objective approaches for trading off performance: maximum likelihood estimation with increasing explicit regularization and maximum a posteriori estimation with varying penalty strength. Results demonstrate that the proposed method generates solutions with comparable to improved objective function values compared to the conventional approaches for trading off performance amongst different tasks. In addition, this approach identifies a diverse set of solutions in the multi-objective function space which can be challenging to estimate with single-objective formulations. [ABSTRACT FROM AUTHOR]