Sequential anomaly detection in the presence of noise and limited feedback

This paper describes a methodology for detecting anomalies from sequentially observed and potentially noisy data. The proposed approach consists of two main elements: 1) filtering, or assigning a belief or likelihood to each successive measurement based upon our ability to predict it from previous noisy observations and 2) hedging, or flagging potential anomalies by comparing the current belief against a time-varying and data-adaptive threshold. The threshold is adjusted based on the available feedback from an end user. Our algorithms, which combine universal prediction with recent work on online convex programming, do not require computing posterior distributions given all current observations and involve simple primal-dual parameter updates. At the heart of the proposed approach lie exponential-family models which can be used in a wide variety of contexts and applications, and which yield methods that achieve sublinear per-round regret against both static and slowly varying product distributions with marginals drawn from the same exponential family. Moreover, the regret against static distributions coincides with the minimax value of the corresponding online strongly convex game. We also prove bounds on the number of mistakes made during the hedging step relative to the best offline choice of the threshold with access to all estimated beliefs and feedback signals. We validate the theory on synthetic data drawn from a time-varying distribution over binary vectors of high dimensionality, as well as on the Enron email dataset. © 1963-2012 IEEE.

Coordinated analysis of delayed sprites with high-speed images and remote electromagnetic fields

Simultaneous measurements of high-altitude optical emissions and magnetic fields produced by sprite-associated lightning discharges enable a close examination of the link between low-altitude lightning processes and high-altitude sprite processes. We report results of the coordinated analysis of high-speed sprite video and wideband magnetic field measurements recorded simultaneously at Yucca Ridge Field Station and Duke University. From June to August 2005, sprites were detected following 67 lightning strokes, all of which had positive polarity. Our data showed that 46% of the 83 discrete sprite events in these sequences initiated more than 10 ms after the lightning return stroke, and we focus on these delayed sprites in this work. All delayed sprites were preceded by continuing current moments that averaged at least 11 kA km between the return stroke and sprites. The total lightning charge moment change at sprite initiation varied from 600 to 18,600 C km, and the minimum value to initiate long-delayed sprites ranged from 600 for 15 ms delay to 2000 C km for more than 120 ms delay. We numerically simulated electric fields at altitudes above these lightning discharges and found that the maximum normalized electric fields are essentially the same as fields that produce short-delayed sprites. Both estimated and simulation-predicted sprite initiation altitudes indicate that long-delayed sprites generally initiate around 5 km lower than short-delayed sprites. The simulation results also reveal that slow (5-20 ms) intensifications in continuing current can play a major role in initiating delayed sprites. Copyright 2008 by the American Geophysical Union.

Broadband chaos generated by an optoelectronic oscillator.

We study an optoelectronic time-delay oscillator that displays high-speed chaotic behavior with a flat, broad power spectrum. The chaotic state coexists with a linearly stable fixed point, which, when subjected to a finite-amplitude perturbation, loses stability initially via a periodic train of ultrafast pulses. We derive approximate mappings that do an excellent job of capturing the observed instability. The oscillator provides a simple device for fundamental studies of time-delay dynamical systems and can be used as a building block for ultrawide-band sensor networks.

Early versus delayed fixed dose combination abacavir/lamivudine/zidovudine in patients with HIV and tuberculosis in Tanzania.

Fixed dose combination abacavir/lamivudine/zidovudine (ABC/3TC/ZDV) among HIV-1 and tuberculosis (TB)-coinfected patients was evaluated and outcomes between early vs. delayed initiation were compared. In a randomized, pilot study conducted in the Kilimanjaro Region of Tanzania, HIV-infected inpatients with smear-positive TB and total lymphocyte count <1200/mm(3) were randomized to initiate ABC/3TC/ZDV either 2 (early) or 8 (delayed) weeks after commencing antituberculosis therapy and were followed for 104 weeks. Of 94 patients screened, 70 enrolled (41% female, median CD4 count 103 cells/mm(3)), and 33 in each group completed 104 weeks. Two deaths and 12 serious adverse events (SAEs) were observed in the early arm vs. one death, one clinical failure, and seven SAEs in the delayed arm (p = 0.6012 for time to first grade 3/4 event, SAE, or death). CD4 cell increases were +331 and +328 cells/mm(3), respectively. TB-immune reconstitution inflammatory syndromes (TB-IRIS) were not observed in any subject. Using intent-to-treat (ITT), missing = failure analyses, 74% (26/35) vs. 89% (31/35) randomized to early vs. delayed therapy had HIV RNA levels <400 copies/ml at 104 weeks (p = 0.2182) and 66% (23/35) vs. 74% (26/35), respectively, had HIV RNA levels <50 copies/ml (p = 0.6026). In an analysis in which switches from ABC/3TC/ZDV = failure, those receiving early therapy were less likely to be suppressed to <400 copies/ml [60% (21/35) vs. 86% (30/35), p = 0.030]. TB-IRIS was not observed among the 70 coinfected subjects beginning antiretroviral treatment. ABC/3TC/ZDV was well tolerated and resulted in steady immunologic improvement. Rates of virologic suppression were similar between early and delayed treatment strategies with triple nucleoside regimens when substitutions were allowed.