A novel unified Bayesian framework for network detection is developed, under which a detection algorithm is derived based on random walks on graphs. The algorithm detects threat networks using partial observations of their activity, and is proved to be optimum in the Neyman-Pearson sense. The algorithm is defined by a graph, at least one observation, and a diffusion model for threat. A link to well-known spectral detection methods is provided, and the equivalence of the random walk and harmonic solutions to the Bayesian formulation is proven. A general diffusion model is introduced that utilizes spatio-temporal relationships between vertices, and is used for a specific space-time formulation that leads to significant performance improvements on coordinated covert networks. This performance is demonstrated using a new hybrid mixed-membership blockmodel introduced to simulate random covert networks with realistic properties.Comment: IEEE Trans. Signal Process., major revision of arxiv.org/abs/1303.5613. arXiv admin note: substantial text overlap with arXiv:1303.561
This paper presents a novel method of using psychoacoustic information from human listening experiments to generate useful features for automated signal classification or regression. The design and analysis of a similarity experiment using active sonar transient echoes is summarized and two methods are presented for feature identification based on the results of the listening experiment. These methods not only identify novel features but also provide a visual insight into perceptually significant signal attributes. The approach presented is based on perceptual similarity measures collected during formal listening experiments but is applicable to any perceptual similarity experiment (e.g. visual).
This paper addresses threat propagation on space-time graphs, defined to be a time-sampled graph. The application considered is geographical sites connected by tracks, though such graphs arise in many fields. Several new concepts and efficient algorithms are introduced, specifically, the space-time adjacency matrix and harmonic threat propagation. The cued threat propagation problem is shown to be equivalent to the harmonic solution to Laplace's equation on the graph. Alternately, the Perron-Frobenius theorem is applied to a modified space-time adjacency matrix to derive a concept of eigen-threat on space-time graphs. Both approaches yield fast, scalable algorithms for space-time threat propagation applicable to both very small and very large graphs. Algorithms are motivated by a continuous time stochastic process model. Detection performance is shown using a simulated insurgent network data for which harmonic space-time threat propagation achieves an 84% probability of detection with a 4% false alarm probability over the entire graph.
The goal of this effort is to develop automatic target classification technology for active sonar systems by exploiting knowledge of signal processing methods and human auditory processing. Using impulsive-source active sonar data, formal listening experiments were conducted to determine if and how human subjects can discriminate between sonar target and clutter echoes using aural cues alone. Both trained sonar operators and naive listeners at APL-UW were examined to determine a baseline performance level. This level was found to be well above chance for multiple subjects in both groups, validating the accepted wisdom that there are inherent aural cues separating targets from clutter. In a subsequent experiment, feedback was provided to the naive listeners and classification performance dramatically improved, demonstrating that naive listeners can be trained to a level on par with experts. Using these trained listeners at APL-UW, a multidimensional scaling (MDS) listening experiment was designed and conducted. The results of these experiments and an analysis of the data, particularly its correlation with the physical attributes of target and clutter echoes (i.e., signal features), will be presented.
Constipation is a common reason for children seeking medical care worldwide. Abdominal complaints and constipation are also common in lead-poisoned children. This study evaluates the prevalence of abnormal blood lead levels (BLL) among pediatric and adolescent patients and examines the association of constipation with elevated BLL. This was a prospective data collection of patients younger than 18 years old with the chief complaint of constipation seen in the Mofid Children’s Hospital gastroenterology clinic and Loghman Hakim pediatric and pediatric gastroenterology clinics were eligible for enrollment in this study. Constipation was defined as infrequent or difficult defecation according to ROME IV criteria lasting 2 months or more. BLL was measured with a fresh capillary whole blood capillary sample. The LeadCare II device assays BLL using an electrochemical technique (anodic stripping voltammetry). A total of 237 patients were enrolled in the study. 122 (51.48%) were female and 115 (48.52%) were male. About one fifth of patients (49; 20.67%) had BLL ≥ 5 µg/dL. The mean BLL in the sample was 3.51 µg/dL. Abdominal pain was the most common symptom accompanying constipation (134; 56%). Multivariate analysis found endoscopic evaluation (P values 0.024, OR 3.646, 95% CI 1.189–11.178), muscle pain (P values 0.020, OR 24.74, 95% CI 1.67–365.83), and maternal education (P values 0.02, OR 4.45, 95% CI 1.27–15.57) with significant differences in groups of patients with normal and elevated BLL. Elevated BLL necessitates an assessment and plans to reduce childhood lead exposure. BLL screening in childhood constipation with refractory chronic abdominal pain may also eradicate the need for invasive procedures like endoscopic evaluation.
Dimensionality reduction algorithms have become an indispensable tool for working with high-dimensional data in classification. Linear discriminant analysis (LDA) is a popular analysis technique used to project high-dimensional data into a lower-dimensional space while maximizing class separability. Although this technique is widely used in many applications, it suffers from overfitting when the number of training examples is on the same order as the dimension of the original data space. When overfitting occurs, the direction of the LDA solution can be dominated by low-energy noise and therefore the solution becomes non-robust to unseen data. In this paper, we propose a novel algorithm, energy-constrained discriminant analysis (ECDA), that overcomes the limitations of LDA by finding lower dimensional projections that maximize inter-class separability, while also preserving signal energy. Our results show that the proposed technique results in higher classification rates when compared to comparable methods. The results are given in terms of SAR image classification, however the algorithm is broadly applicable and can be generalized to any classification problem.
In many aural (acoustic) signal processing tasks, humans are known to perform better than automated classification systems. For such applications, it may be beneficial to identify how humans relate different sounds to one another and incorporate that information into an automatic classification system. This paper presents a method for using psychoacoustic information from a human listening experiment to generate a novel kernel function that can be used to improve automated signal classification. We have conducted a similarity-based listening experiment on a series of impulsive-source sonar echoes. In this experiment, humans were asked to rate perceived similarity between pairings of target and clutter echoes. These ratings combine to form a similarity matrix that reflects the underlying distance measure humans use when judging these echoes. This similarity matrix is a perceptual equivalent to the similarity matrix used in modern kernel methods used in automatic classification systems (e.g. Support Vector Machine). By fitting an appropriate distance metric to the results of the perceptual experiment we can identify novel, perceptually-inspired, kernel functions. This paper presents a series of new approaches for the identification of a perceptual kernel function. We then compare the classification performance between these perceptual kernels and more standard kernel functions.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.