Kenneth W. Bauer

Feature saliency measures

Abstract This paper presents a survey of feature saliency measures used in artificial neural networks. Saliency measures can be used for assessing a features relative importance. In this paper, we contrast two basic philosophies for measuring feature saliency or importance within a feed-forward neural network. One philosophy is to evaluate each feature with respect to relative changes in either the neural networks output or the neural networks probability of error. We refer to this as a derivative-based philosophy of feature saliency. Using the derivative-based philosophy, we propose a new and more efficient probability of error measure. A second philosophy is to measure the relative size of the weight vector emanating from each feature. We refer to this as a weight-based philosophy of feature saliency. We derive several unifying relationships which exist within the derivative-based feature saliency measures, as well as between the derivative and the weight-based feature saliency measures. We also report experimental results for an target recognition problem using a number of derivative-based and weight-based saliency measures.

Improving pilot mental workload classification through feature exploitation and combination

Pred icting high pilot mental workloadis important to the UnitedStates Air Force because lives andaircraft have been lost due to errors made during periods of ;ight associated with mental overload and task saturation. Current research e0orts use psychophysiological measures such as electroencephalography (EEG), cardiac, ocular, and respiration measures in an attempt to identify and predict mental workload levels. Existing classi

Finding Hyperspectral Anomalies Using Multivariate Outlier Detection

cation methods successfully classify pilot mental workload using ;ight data for a single pilot on a given day, but are unsuccessful across di0erent pilots and/or days. We demonstrate a small subset of combinedandcalibratedpsychophysiological features collectedfrom a single pilot on a given d ay that accurately classi

Three-dimensional receiver operating characteristic (ROC) trajectory concepts for the evaluation of target recognition algorithms faced with the unknown target detection problem

es mental workload for a separate pilot on a di0erent day. We achieve classi

A locally adaptable iterative RX detector

cation accuracy (CA) improvements over previous classi

Simulation of stochastic activity networks using path control variates

ers exceeding 80% while using signi

Response surface methodology as a sensitivity analysis tool in decision analysis

cantly fewer features and dramatically reducing the CA variance. Without the need for EEG data, our feature combination and calibration scheme also radically reduces the raw data collection requirements, making data collection immensely easier to manage andspectacularly red ucing computational processing requirements. ? 2004 Publishedby Elsevier Ltd .

Feature Selection for RF Fingerprinting With Multiple Discriminant Analysis and Using ZigBee Device Emissions

This research demonstrates the adverse implications of using non-robust statistical methods for detecting anomalies in hyperspectral image data, and proposes the use of multivariate outlier detection methods as an alternative detection strategy. Existing outlier detection methods are adapted for use in a hyperspectral image context, and their performance is compared to the benchmark RX detector and a cluster-based anomaly detector. Tests conducted using both simulated data and actual hyperspectral imagery indicate that multivariate outlier detection methods can achieve superior detection performance relative to current non-robust detection methods.

Principal Component Reconstruction Error for Hyperspectral Anomaly Detection

This paper presents a three-dimensional (3-D) receiver operating characteristic (ROC) trajectory which combines the principles of the standard ROC curve used in automatic target recognition (ATR) with a third performance measure. These 3-D ROC trajectories are used to compare competing target recognition algorithms when unknown targets are present in the data. A probability of rejection is added to the standard ROC curve that includes the probability of successful detection (true positive rate) and the probability of false alarm (false positive rate). By using the probability of rejection, targets may be eliminated that are difficult to classify, or perhaps unknown. The 3-D ROC trajectory can be a useful tool for a SAR image analyst for understanding the tradeoffs between the probability of rejection and the two standard performance measures commonly used in detection problems.

AutoGAD: An Improved ICA-Based Hyperspectral Anomaly Detection Algorithm

We present an unsupervised anomaly detection method for hyperspectral imagery (HSI) based on data characteristics inherit in HSI. A locally adaptive technique of iteratively refining the well-known RX detector (LAIRX) is developed. The technique is motivated by the need for better first- and second-order statistic estimation via avoidance of anomaly presence. Overall, experiments show favorable Receiver Operating Characteristic (ROC) curves when compared to a global anomaly detector based upon the Support Vector Data Description (SVDD) algorithm, the conventional RX detector, and decomposed versions of the LAIRX detector. Furthermore, the utilization of parallel and distributed processing allows fast processing time making LAIRX applicable in an operational setting.