Time filter

Source Type

Dachman-Soled D.,University of Maryland College Park | Feldman V.,IBM | Tan L.-Y.,University of California at Berkeley | Wan A.,Institute for Defense Analysis | Wimmer K.,Duquesne University
Proceedings of the Annual ACM-SIAM Symposium on Discrete Algorithms

A function f is d-resilient if all its Fourier coefficients of degree at most d are zero, i.e. f is uncorrected with all low-degree parities. We study the notion of approximate resilience of Boolean functions, where we say that f is α-approximately d-resilient if /is α-close to a [-1, l]-valued d-resilient function in ℓ1 distance. We show that approximate resilience essentially characterizes the complexity of agnostic learning of a concept class C over the uniform distribution. Roughly speaking, if all functions in a class C are far from being d-resilient then C can be learned agnostically in time no(d) and conversely, if C contains a function close to being d-resilient then agnostic learning of C in the statistical query (SQ) framework of Kearns has complexity of at least nω{d). Focusing on monotone Boolean functions, we exhibit the existence of near-optimal α-approximately ω(a√n)-resilient monotone functions for all α > 0. Prior to our work, it was conceivable even that every monotone function is ω(1)-far from any 1-resilient function. Furthermore, we construct simple, explicit monotone functions based on Tribes and CycleRun that are close to highly resilient functions. Our constructions are based on general resilience analysis and amplification techniques we introduce. These structural results, together with the characterization, imply nearly optimal lower bounds for agnostic learning of monotone juntas, a natural variant of the well-studied junta learning problem. In particular we show that no SQ algorithm can efficiently agnostically learn monotone κ-juntas for any κ = ω(1) and any constant error less than 1/2. Copyright © 2015 by the Society for Industrial and Applied Mathmatics. Source

Bieringer P.E.,U.S. National Center for Atmospheric Research | Annunzio A.J.,U.S. National Center for Atmospheric Research | Platt N.,Institute for Defense Analysis | Bieberbach G.,U.S. National Center for Atmospheric Research | Hannan J.,Defense Threat Reduction Agency
Journal of Applied Meteorology and Climatology

Chemical and biological (CB) defense systems require significant testing and evaluation before they are deployed for real-time use. Because it is not feasible to evaluate these systems with open-air testing alone, researchers rely on numerical models to supplement the defense-system analysis process. These numerical models traditionally describe the statistical properties of CB-agent atmospheric transport and dispersion (AT&D). While the statistical representation of AT&D is appropriate to use in some CB defense analyses, it is not appropriate to use this class of dispersion model for all such analyses. Many of these defense-system analyses require AT&D models that are capable of simulating dispersion properties with very short timeaveraging periods that more closely emulate a ''single realization'' of a contaminant or CB agent dispersing in a turbulent atmosphere. The latter class ofAT&Dmodels is superior to the former for performing CB-system analyses when one or more of the following factors are important in the analysis: high-frequency sampling of the contaminant, spatial and temporal correlations within the contaminant concentration field, and nonlinear operations performed on the contaminant concentration. This paper describes and contrasts these AT&D modeling tools and provides specific examples in which utilizing ensembles of single realizations of CB-agent AT&D is advantageous over using the statistical, ''ensemble-average'' representation of the agent AT&D. These examples demonstrate the importance of using an AT&D modeling tool that is appropriate for the analysis. © 2014 American Meteorological Society. Source

Chang J.,Princeton University | Fok M.P.,Princeton University | Fok M.P.,Lightwave Photonics, Inc. | Meister J.,Institute for Defense Analysis | Prucnal P.R.,Princeton University
Optics Express

In this paper we present a fully tunable and reconfigurable single-laser multi-tap microwave photonic FIR filter that utilizes a special SM-to-MM combiner to sum the taps. The filter requires only a single laser source for all the taps and a passive component, a SM-to-MM combiner, for incoherent summing of signal. The SM-to-MM combiner does not produce optical interference during signal merging and is phase-insensitive. We experimentally demonstrate an eight-tap filter with both positive and negative programmable coefficients with excellent correspondence between predicted and measured values. The magnitude response shows a clean and accurate function across the entire bandwidth, and proves successful operation of the FIR filter using a SM-to-MM combiner. © 2013 Optical Society of America. Source

Seda-Sanabria Y.,U.S. Army | Matheu E.E.,Critical Lifelines Branch | Morgeson J.D.,Institute for Defense Analysis | Kirpichevsky Y.,Institute for Defense Analysis | And 3 more authors.
International Journal on Hydropower and Dams

The Common Risk Model for Dams (CRM-D), described here, was developed as a result of collaboration between the US Army Corps of Engineers and the US Department of Homeland Security. It is used for security risk assessment of dams, navigation locks, hydro projects, and similar infrastructure. The method provides a systematic approach for evaluating and comparing security risks across a large portfolio. Risk is calculated for attack scenarios (specific adversary using a specific attack vector against a specific target) by combining consequence, vulnerability, and threat estimates in a way that accounts for the relationships among these variables. The CRM-D can effectively quantify the benefits of implementing a particular risk mitigation strategy and, consequently, enable return-on-investment analyses for multiple mitigation options across a large portfolio. Source

Platt N.,Institute for Defense Analysis | Czech C.,Clemson University | Urban J.,Institute for Defense Analysis | Deriggi D.,Institute for Defense Analysis | And 5 more authors.
HARMO 2011 - Proceedings of the 14th International Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes

The most common way for an atmospheric transport and dispersion (AT&D) model to calculate toxic effects is based on the total inhaled dose. These effects are independent of the manner in which this dose was accumulated. But for many chemicals, it has been observed that the time dependence of the concentration is important. Toxic load modelling tries to account for this by utilizing the toxic load exponent. While the experimental data supporting toxic load modelling were derived using a rectangular pulse concentration exposure profiles, the actual exposures from hazardous plumes are temporally highly variable. There are several proposed generalizations of the toxic load model to the case of time-varying exposure profiles, none of which have been validated using animal experiments. The majority of present-day AT&D models used for consequence assessment estimate a "mean" plume that approximates the ensemble average over a large number of plume realizations. The few models that utilize the toxic load model apply it using this mean plume. We have obtained a high resolution (in space and time) set of VTHREAT predictions that contains 20 plume realizations for a continuous release of a neutral-buoyancy tracer gas over flat terrain under stable atmospheric conditions. We use this data set for comparing toxic load calculations based on a "mean" plume with calculations based on individual plume realizations in order to assess the potential effect on casualty estimation. Source

Discover hidden collaborations