We address the problem of detecting slow-moving targets using space-time adaptive processing (STAP) radar. Determining the optimum weights at each range requires data snapshots at neighboring ranges. However, in virtually all configurations, snapshot statistics are range dependent, meaning that snapshots are nonstationary with respect to range. This results in poor performance. In this paper, w...

We introduce a novel, greatly simplified classifier for binarized data. The model contains a sparse, “digital” hidden layer of Parity interactions, followed by a sigmoidal output node. We propose priors for the cases: a) input space obeys a metrics, b) inputs encode discrete attributes. Stochastic search for the hidden layer allows capacity and smoothness of the approximation to be controlled b...

We address the problem of detecting slowmoving targets using space-time adaptive processing (STAP). The construction of the optimum weights at each range implies the estimation of the clutter covariance matrix. This is typically done by straight averaging of snapshots at neighboring ranges. However, in most configurations, the snapshots’ statistics are range-dependent. Straight averaging thus r...

We propose a model for non-preemptive scheduling under uncertainty. In this model, we combine the main characteristics of online and stochastic scheduling in a simple and natural way. Job processing times are assumed to be stochastic, but in contrast to the traditional stochastic scheduling models, we assume that jobs arrive online over time, and there is no knowledge about the jobs that will a...

We present and compare two delayed random walk models of neural information processing. A delayed random walk is a random walk in which the transition probability depends on the position of the walker at a time in the past. The rst model is of human neuro-muscular control for postural sway. The other model is for a stochastic single neuron which has a delayed self{exciting feedback. We study ea...

This paper presents a novel conditionally suboptimal filtering algorithm on estimation problems that arise in discrete nonlinear time-varying stochastic difference systems. The suboptimal state estimate is formed by summing of conditionally nonlinear filtering estimates that their weights depend only on time instants, in contrast to conditionally optimal filtering, the proposed conditionally su...

Let {Yt} be an observed time series where the interval between observations is fixed. We consider models which take a window of length d of the time series so that Ys+d is the response to 〈Ys, Ys+1, . . . , Ys+d−1〉. We will train the models to predict the next observation in the time series. In order to predict an observation some number of periods f in the future we will predict the next obser...

Standard time series analysis estimates the power spectral density over the full frequency range, until half the sampling frequency. In several input–output identification problems, frequency selective model estimation is desirable. Processing of a time series in a subband may also be useful if observations of a stochastic process are analyzed for the presence or multiplicity of spectral peaks....

Many analyses of visual search assume error-free component processes. These analyses range from Sternberg's earliest serial models to Townsend's sophisticated theorems of serial-parallel equivalence. Consider a simple "yes-no" visual search task with a set-size manipulation. For the correct positive responses, an error-free, unlimited-capacity, parallel search model predicts no effect of set si...

For variable selection to binary response regression, stochastic search variable selection and Bayesian Lasso have recently been popular. However, these two variable selection methods suffer from heavy computation burden caused by hyperparameter tuning and by matrix inversions, especially when the number of covariates is large. Therefore, this article incorporates the componenetwise Gibbs sampl...