Publications

Refine Results

(Filters Applied) Clear All

Subgraph Detection

Author:
Published in:
Graph Algorithms in the Language of Linear Algebra, pp. 115-133.

Summary

Detecting subgraphs of interest in larger graphs is the goal of many graph analysis techniques. The basis of detection theory is computing the probability of a “foreground” with respect to a model of the “background” data. Hidden Markov Models represent one possible foreground model for patterns of interaction in a graph. Likewise, Kronecker graphs are one possible model for power law background graphs. Combining these models allows estimates of the signal to noise ratio, probability of detection, and probability of false alarm for different classes of vertices in the foreground. These estimates can then be used to construct filters for computing the probability that a background graph contains a particular foreground graph. This approach is applied to the problem of detecting a partially labeled tree graph in a power law background graph. One feature of this method is the ability to a priori estimate the number of vertices that will be detected via the filter.
READ LESS

Summary

Detecting subgraphs of interest in larger graphs is the goal of many graph analysis techniques. The basis of detection theory is computing the probability of a “foreground” with respect to a model of the “background” data. Hidden Markov Models represent one possible foreground model for patterns of interaction in a...

READ MORE

Linear algebraic notation and definitions

Published in:
Graph Algorithms in the Language of Linear Algebra, pp. 13-18.

Summary

This chapter presents notation, definitions, and conventions for graphs, matrices, arrays, and operations upon them.
READ LESS

Summary

This chapter presents notation, definitions, and conventions for graphs, matrices, arrays, and operations upon them.

READ MORE

A knowledge-based operator for a genetic algorithm which optimizes the distribution of sparse matrix data

Published in:
Parallel Architectures and Bioinspired Algorithms

Summary

We present the Hogs and Slackers genetic algorithm (GA) which addresses the problem of improving the parallelization efficiency of sparse matrix computations by optimally distributing blocks of matrices data. The performance of a distribution is sensitive to the non-zero patterns in the data, the algorithm, and the hardware architecture. In a candidate distributions the Hogs and Slackers GA identifies processors with many operations – hogs, and processors with fewer operations – slackers. Its intelligent operation-balancing mutation operator then swaps data blocks between hogs and slackers to explore a new data distribution.We show that the Hogs and Slackers GA performs better than a baseline GA. We demonstrate Hogs and Slackers GA’s optimization capability with an architecture study of varied network and memory bandwidth and latency.
READ LESS

Summary

We present the Hogs and Slackers genetic algorithm (GA) which addresses the problem of improving the parallelization efficiency of sparse matrix computations by optimally distributing blocks of matrices data. The performance of a distribution is sensitive to the non-zero patterns in the data, the algorithm, and the hardware architecture. In...

READ MORE

Graphs and matrices

Author:
Published in:
Graph Algorithms in the Language of Linear Algebra, pp. 3-12

Summary

A linear algebraic approach to graph algorithms that exploits the sparse adjacency matrix representation of graphs can provide a variety of benefits. These benefits include syntactic simplicity, easier implementation, and higher performance. Selected examples are presented illustrating these benefits. These examples are drawn from the remainder of the book in the areas of algorithms, data analysis, and computation.
READ LESS

Summary

A linear algebraic approach to graph algorithms that exploits the sparse adjacency matrix representation of graphs can provide a variety of benefits. These benefits include syntactic simplicity, easier implementation, and higher performance. Selected examples are presented illustrating these benefits. These examples are drawn from the remainder of the book in...

READ MORE

Retroreflectors for remote readout of colorimetric sensors

Published in:
Sensors and Actuators B-Chemical, Vol. 160, No. 1, 15 December 2011, pp. 1244-1249.

Summary

We have developed a remote detection system consisting of commercially available retroreflective material coated with an analyte-specific colorimetric dye. Quantitative performance modeling predicts that, given the appropriate indicator dye, a system with a 10 cm optic and eye-safe illumination should be capable of detecting small droplets of contamination at kilometer ranges. We have synthesized new colorimetric dyes specific to organophosphate contamination and, with these dyes, demonstrated detection of 1um of liquid malathion at over 150 m with less than 20 mW of laser illumination.
READ LESS

Summary

We have developed a remote detection system consisting of commercially available retroreflective material coated with an analyte-specific colorimetric dye. Quantitative performance modeling predicts that, given the appropriate indicator dye, a system with a 10 cm optic and eye-safe illumination should be capable of detecting small droplets of contamination at kilometer...

READ MORE

Topic modeling for spoken documents using only phonetic information

Published in:
ASRU 2011, IEEE Workshop on Automatic Speech Recognition & Understanding, 11-15 December 2011, pp. 395-400.

Summary

This paper explores both supervised and unsupervised topic modeling for spoken audio documents using only phonetic information. In cases where word-based recognition is unavailable or infeasible, phonetic information can be used to indirectly learn and capture information provided by topically relevant lexical items. In some situations, a lack of transcribed data can prevent supervised training of a same-language phonetic recognition system. In these cases, phonetic recognition can use cross-language models or self-organizing units (SOUs) learned in a completely unsupervised fashion. This paper presents recent improvements in topic modeling using only phonetic information. We present new results using recently developed techniques for discriminative training for topic identification used in conjunction with recent improvements in SOU learning. A preliminary examination of the use of unsupervised latent topic modeling for unsupervised discovery of topics and topically relevant lexical items from phonetic information is also presented.
READ LESS

Summary

This paper explores both supervised and unsupervised topic modeling for spoken audio documents using only phonetic information. In cases where word-based recognition is unavailable or infeasible, phonetic information can be used to indirectly learn and capture information provided by topically relevant lexical items. In some situations, a lack of transcribed...

READ MORE

Radiation effects in 3D integrated SOI SRAM circuits

Summary

Radiation effects are presented for the first time for vertically integrated 3 x 64 -kb SOI SRAM circuits fabricated using the 3D process developed at MIT Lincoln Laboratory. Three fully-fabricated 2D circuit wafers are stacked using standard CMOS fabrication techniques including thin-film planarization, layer alignment and oxide bonding. Micron-scale dense 3D vias are fabricated to interconnect circuits between tiers. Ionizing dose and single event effects are discussed for proton irradiation with energies between 4.8 and 500 MeV. Results are compared with 14-MeV neutron irradiation. Single event upset cross section, tier-to-tier and angular effects are discussed. The interaction of 500-MeV protons with tungsten interconnects is investigated usingMonte-Carlo simulations. Results show no tier-to-tier effects and comparable radiation effects on 2D and 3D SRAMs. 3DIC technology should be a good candidate for fabricating circuits for space applications.
READ LESS

Summary

Radiation effects are presented for the first time for vertically integrated 3 x 64 -kb SOI SRAM circuits fabricated using the 3D process developed at MIT Lincoln Laboratory. Three fully-fabricated 2D circuit wafers are stacked using standard CMOS fabrication techniques including thin-film planarization, layer alignment and oxide bonding. Micron-scale dense...

READ MORE

SET characterization in logic circuits fabricated in a 3DIC technology

Summary

Single event transients are characterized for the first time in logic gate circuits fabricated in a novel 3DIC technology where SET test circuits are vertically integrated on three tiers in a 20-um-thick layer. This 3D technology is extremely will suited for high-density circuit integration because of the small dimension the tier-to-tier circuit interconnects, which are 1.25-um-wide-through-oxide-vias. Transient pulse width distributions were characterized simultaneously on each tier during exposure to krypton heavy ions. The difference in SET pulse width and cross-section among the three tiers is discussed. Experimental test results are explaine dby considering the electrical characteristics of the FETs on the 2D wafers before 3D integration, and by considering the energy deposited by the Kr ions passing through the various material laters of the 3DIC stack. We also show that the backmetal layer available on the upper tiers can be used to tune independently the nFET and pFET current drive, and change the SET pulse width and cross-section. This 3DIC technology appears to be a good candidate for space applications.
READ LESS

Summary

Single event transients are characterized for the first time in logic gate circuits fabricated in a novel 3DIC technology where SET test circuits are vertically integrated on three tiers in a 20-um-thick layer. This 3D technology is extremely will suited for high-density circuit integration because of the small dimension the...

READ MORE

Investigating acoustic correlates of human vocal fold vibratory phase asymmetry through modeling and laryngeal high-speed videoendoscopy

Published in:
J. Acoust. Soc. Am., Vol. 130, No. 6, December 2011, pp. 3999-4009.

Summary

Vocal fold vibratory asymmetry is often associated with inefficient sound production through its impact on source spectral tilt. This association is investigated in both a computational voice production model and a group of 47 human subjects. The model provides indirect control over the degree of left-right phase asymmetry within a nonlinear source-filter framework, and high-speed videoendoscopy provides in vivo measures of vocal fold vibratory asymmetry. Source spectral tilt measures are estimated from the inverse-filtered spectrum of the simulated and recorded radiated acoustic pressure. As expected, model simulations indicated that increasing left-right phase asymmetry induces steeper spectral tilt. Subject data, however, reveal that none of the vibratory asymmetry measures correlates with spectral tilt measures. Probing further into physiological correlates of spectral tilt that might be affected by asymmetry, the glottal area waveform is parameterized to obtain measures of the open phase (open/plateau quotient) and closing phase (speed/closing quotient). Subjects' left-right phase asymmetry exhibits low, but statistically significant, correlations with speed quotient (r=0.45) and closing quotient (r=-0.39). Results call for future studies into the effect of asymmetric vocal fold vibrartion on glottal airflow and the associated impact on voice source spectral properties and vocal efficiency.
READ LESS

Summary

Vocal fold vibratory asymmetry is often associated with inefficient sound production through its impact on source spectral tilt. This association is investigated in both a computational voice production model and a group of 47 human subjects. The model provides indirect control over the degree of left-right phase asymmetry within a...

READ MORE

Exploring the variable sky with LINEAR : photometric recalibration with the Sloan Digital Sky Survey

Published in:
Astronomical J., Vol. 142, No. 6, December 2011.

Summary

We describe photometric recalibration of data obtained by the asteroid survey LINEAR. Although LINEAR was designed for astrometric discovery of moving objects, the data set described here contains over 5 billion photometric measurements for about 25 million objects, mostly stars. We use Sloan Digital Sky Survey (SDSS) data from the overlapping ~10,000 deg2 of sky to recalibrate LINEAR photometry and achieve errors of 0.03 mag for sources not limited by photon statistics with errors of 0.2 mag at r ~ 18. With its 200 observations per object on average, LINEAR data provide time domain information for the brightest four magnitudes of the SDSS survey. At the same time, LINEAR extends the deepest similar wide-area variability survey, the Northern Sky Variability Survey, by 3 mag.We briefly discuss the properties of about 7000 visually confirmed periodic variables, dominated by roughly equal fractions of RR Lyrae stars and eclipsing binary stars, and analyze their distribution in optical and infrared color?color diagrams. The LINEAR data set is publicly available from the SkyDOT Web site.
READ LESS

Summary

We describe photometric recalibration of data obtained by the asteroid survey LINEAR. Although LINEAR was designed for astrometric discovery of moving objects, the data set described here contains over 5 billion photometric measurements for about 25 million objects, mostly stars. We use Sloan Digital Sky Survey (SDSS) data from the...

READ MORE