Data strip mining for the virtual design of pharmaceuticals withneural networks

A novel neural network based technique, called "data strip mining" extracts predictive models from data sets which have a large number of potential inputs and comparatively few data points. This methodology uses neural network sensitivity analysis to determine which predictors are most significant in the problem. Neural network sensitivity analysis holds all but one input to a trained neural network constant while varying each input over its entire range to determine its effect on the output. Elimination of variables through neural network sensitivity analysis and predicting performance through model cross-validation allows the analyst to reduce the number of inputs and improve the model's predictive ability at the same time. This paper demonstrates its effectiveness on a pair of problems from combinatorial chemistry with over 400 potential inputs each. For these data sets, model selection by neural sensitivity analysis outperformed other variable selection methods including the forward selection and genetic algorithm.     

DARPA Information Assurance Program dynamic defense experimentsummary

Several types of experiments are being conducted by the Defense Advanced Research Projects Agency (DARPA) Information Assurance (IA) Program in DARPA's IA Lab. This research program is driven by concepts of strategic cyberdefense. Each experiment involves a carefully formulated hypothesis that is intended to be either supported or refuted by the experimental testing. In many cases, “red team” attackers participate in all phases of the experiment and contribute to generating the data required to test the hypothesis. The red team is usually structured to model a well-resourced adversary, such as a foreign, national intelligence agency. The particular experiment described here explored one aspect of the IA program's grand hypothesis of dynamic defense: “Dynamic modification of defensive structure improves system assurance.” This experiment concentrated on the assertion that autonomic response mechanisms can improve overall system assurance by thwarting an attack while it is underway. In most cases, each attack in this experiment was run first with only “prevent and detect” mechanisms enabled, then repeated with “prevent, detect, and respond mechanisms” enabled. The key result of this experiment is that the hypothesis was supported     

Improving energy resolution of calorimeters using a covariance matrix approach

A covariance matric technique is used to improve energy resolution of electromagnetic and hadronic showers produced in a uranium/liquid argon calorimeter. An event of unknown energy is compared with parametrized energy depositions through a χ² minimization procedure, thereby providing an estimate of the true energy of the shower. The method is effective for calorimeters that have large amounts of uninstrumented material.     

Validation of sensor alert correlators

The authors describe the first experimental validation of correlation systems with the goal of assessing the overall progress in the field. Their experiment set out to measure the collective ability of correlators to recognize cyber attacks and designate their targets.     

Computational military tactical planning system

A computational system called fuzzy-genetic decision optimization combines two soft computing methods, genetic optimization and fuzzy ordinal preference, and a traditional hard computing method, stochastic system simulation, to tackle the difficult task of generating battle plans for military tactical forces. Planning for a tactical military battle is a complex, high-dimensional task which often bedevils experienced professionals. In fuzzy-genetic decision optimization, the military commander enters his battle outcome preferences into a user interface to generate a fuzzy ordinal preference model that scores his preference for any battle outcome. A genetic algorithm iteratively generates populations of battle plans for evaluation in a stochastic combat simulation. The fuzzy preference model converts the simulation results into a fitness value for each population member, allowing the genetic algorithm to generate the next population. Evolution continues until the system produces a final population of high-performance plans which achieve the commander's intent for the mission. Analysis of experimental results shows that co-evolution of friendly and enemy plans by competing genetic algorithms improves the performance of the planning system. If allowed to evolve long enough, the plans produced by automated algorithms had a significantly higher mean performance than those generated by experienced military experts.