Abraham Schultz

Collective recall via the brain-state-in-a-box network

A number of approaches to pattern recognition employ variants of nearest neighbor recall. This procedure uses a number of prototypes of known class and identifies an unknown pattern vector according to the prototype it is nearest to. A recall criterion of this type that depends on the relation of the unknown to a single prototype is a non-smooth function and leads to a decision boundary that is a jagged, piecewise linear hypersurface. Collective recall, a pattern recognition method based on a smooth nearness measure of the unknown to all the prototypes, is developed. The prototypes are represented as cells in a brain-state-in-a-box (BSB) network. Cells that represent the same pattern class are linked by positive weights and cells representing different pattern classes are linked by negative weights. Computer simulations of collective recall used in conjunction with learning vector quantization (LVQ) show significant improvement in performance relative to nearest neighbor recall for pattern classes defined by nonspherically symmetric Gaussians.

Morphology and autowave metric on CNN applied to bubble-debris classification

In this study, we present the initial results of cellular neural network (CNN)-based autowave metric to high-speed pattern recognition of gray-scale images. the application is to a problem involving separation of metallic wear debris particles from air bubbles. This problem arises in an optical-based system for determination of mechanical wear. This paper focuses on distinguishing debris particles suspended in the oil flow from air bubbles and aims to employ CNN technology to create an online fault monitoring system. For the class of engines of interest bubbles occur much more often than debris particles and the goal is to develop a classification system with an extremely low false alarm rate for misclassified bubbles. The designed analogic CNN algorithm detects and classifies single bubbles es and bubble groups using binary morphology and autowave metric. The debris particles are separated based on autowave distances computed between bubble models and the unknown objects. Initial experiments indicate that the proposed algorithm is robust and noise tolerant and when implemented on a CNN universal chip it provides a solution in real time.

Unsupervised BCM projection pursuit algorithms for classification of simulated radar presentations

Abstract A comparison of the unsupervised Projection Pursuit learning algorithm (BCM), with supervised backward propagation (BP) and a laterally inhibited version of BP (LIBP) was performed. Simulated inverse synthetic aperature radar (ISAR) presentations served as a testbed for evaluation. Symmetries of the artificial presentations make the use of localized moments a convenient preprocessing tool for the inputs. Although all three algorithms obtain classification rates comparable to trained human observers for this simulated data base, BCM obtains solutions that classify more effectively inputs that are corrupted by noise or errors in registration; in noise tolerance experiments, the best BCM solution represents a 10 dB improvement over the best BP solution. Recurrent and differential forms of BCM that could be applied to time-dependent classification problems are also developed.

Optical classification of bioaerosols using UV fluorescence and IR absorption spectroscopy

A partnership that includes the Naval Research Laboratory (NRL), MIT Lincoln Laboratories and the Edgewood Chemical and Biological Command is engaged in an effort to develop optical techniques for the rapid detection and classification of biological aerosols. This paper will describe two efforts at NRL: development of an improved UV fluorescence front-end trigger and the use of infrared absorption spectroscopy to classify biological aerosol particles. UV Laser-induced fluorescence (UVLIF) has been demonstrated to provide very high sensitivity for differentiating between biological and inorganic aerosol particles. Unfortunately, current UVLIF systems have unacceptably high false alarm rates due to interferences from man made and naturally occurring organic and biological particulates. We have developed a two-wavelength, UVLIF technique that offers a higher level of discrimination than is possible using single wavelength UVLIF. Infrared absorption spectroscopy coupled with multivariate analysis demonstrates a high potential for differentiation among members of biological and chemical sample classes. Two-wavelength UVLIF in combination with the IR interrogation of collected bioaerosols could provide a rapid, reagentless approach to specific classification of biological particles according to an operational level of discrimination - the degree of particle characterization required in order to signal the presence of pathogenic material.

Field test results and ambient aerosol measurements using dual wavelength fluorescence excitation and elastic scatter for bioaerosols

A bioaerosol sensor based on dual wavelength fluorescence excitation and multiple wavelength elastic scattering has been developed and characterized for classifying micron-sized particles on the fly. The UVLIF instrument successfully completed a field trial in which we detected and correctly identified over 90% of the simulant releases over the 2 week testing period.

Data fusion in neural networks via computational evolution

Pattern recognition systems commonly employ a single representation of the sensor data. For hard classification problems it is unlikely that a single representation will be able to capture all the relevant information in the sensor field. For a given input, the goal is to fuse the information contained in multiple representations to compute the associated pattern class. For each representation, the learning vector quantization network is first used to establish a transformation to an associated feature space. A recurrent network is then used to fuse the information generated by each of the representations. The weights for the recurrent network are learned using an evolutionary strategy. This network is multi-stable and its equilibrium states are associated with different pattern classes. For a specified input, the system relaxes to an equilibrium state associated with an underlying pattern class. The class decision boundaries generated by the recurrent neural network are compared to the boundaries generated by nearest neighbor recall.<<ETX>>

Classification of simulated radar imagery using lateral inhibition neural networks

The use of neural networks for the classification of simulated inverse synthetic aperture radar imagery is investigated. Symmetries of the artificial imagery make the use of localized moments a convenient preprocessing tool for the inputs to a neural network. A database of simulated targets was obtained by warping dynamical models to representative angles and generating images with differing target motions. Ordinary backward propagation (BP) and some variants of BP which incorporate lateral inhibition (LIBP) obtain a generalization rate of up to approximately 77% for novel data not used during training, a rate which is comparable to the mean level of classification accuracy that trained human observers obtained from the unprocessed simulated imagery. The authors also describe preliminary results for an unsupervised lateral inhibition network based on the BCM neuron. The feature vectors found by BCM are qualitatively different from those of BP and LIBP.<<ETX>>

A discrete dynamics model for synchronization of pulse-coupled oscillators

Biological information processing systems employ a variety of feature types. It has been postulated that oscillator synchronization is the mechanism for binding these features together to realize coherent perception. A discrete dynamic model of a coupled system of oscillators is presented. The network of oscillators converges to a state where subpopulations of cells become phase synchronized. It has potential applications to describing biological perception as well as for the construction of multifeature pattern recognition systems. It is shown that this model can be used to detect the presence of short line segments in the boundary contour of an object. The Hough transform, which is the standard method for detecting curve segments of a specified shape in an image was found not to be effective for this application. Implementation of the discrete dynamics model of oscillator synchronization is much easier than the differential equation models that have appeared in the literature. A systematic numerical investigation of the convergence properties of the model has been performed and it is shown that the discrete dynamics model can scale up to large number of oscillators.

A multiple population Boltzmann machine

Boltzmann machines and genetic algorithms have been successfully applied to function optimization problems. The model developed, merges these approaches to obtain a system that has the best features of both. The composite system offers capabilities difficult to obtain with standard genetic algorithms. It yields automatic niche formation and at the same time it avoids premature convergence. It does not have the Boltzmann machines problem of getting trapped in a local maxima. The model has a temperature parameter that can be used to obtain convergence to a global optimum as is done for simulated annealing. The single population Boltzmann machine is extended to a multiple population and an associated set of genetic operators. It is shown that the equilibrium probability distribution is Gibbs. Computer simulations that show niche formation are presented.<<ETX>>

A Novel Polarized Elastic Scatter Detection Method of Aerosol Particle Velocimetry with Reduced Errors Due to Coincidence and Phantom Particles

This article describes a novel polarized elastic scattering method to measure flow velocities of aerosol particles. Velocities of individual aerosol particles of mixed size, and composition have been determined from time differences between scattered light pulses as they move between two parallel laser beams separated by a fixed distance, known as a time-of-flight method. By using two orthogonally polarized laser beams, the detected light pulses can be assigned to each beam by simultaneously determining their respective depolarization ratio. This implementation of time-of-flight velocimetry provides a mechanically robust distance between the two beams, which reduces error in the measurement and improves both its precision and accuracy, but more importantly, the use of the polarization features of the scattered light reduces errors due to particle coincidence and phantom particles caused by confusion of start and stop timing pulses. This approach was developed and demonstrated in an application to provide effective timing for subsequent selective actions on individual particles downstream such as: further diagnostic measurement, electrical charging and capture. These downstream actions require accurately predicted individual particle trajectories. Copyright 2013 American Association for Aerosol Research