Ranking and Selection

Automated adversarial detection systems can fail when under attack by adversaries. As part of a re- silient data stream mining system to reduce the pos- sibility of such failure, adaptive spike detection is attribute ranking and selection without class-labels. The first part of adaptive spike detection requires weighing all attributes for spiky-ness to rank them. The second part involves filtering

Most model selection mechanisms work in an 'overall' modus, providing models without speciffic concern for how the selected model is going to be used afterwards. The focussed information criterion (FIC), on the other hand, is geared towards optimum model selection when inference is required for a given estimand. In this paper the FIC method is extended to weighted versions. This allows one to rank and select candidate models for the purpose of handling a range of similar tasks well, as opposed to being forced to focus on each task separately. Applications include selecting regression models that perform well for speciffied regions of covariate values. We derive these wFIC criteria, give asymptotic results, and apply the method store al data. Formulae for easy implementation are provided for the class of generalised linear models.

We describe a methodology for performing variable ranking and selection using support vector machines (SVMs). The method constructs a series of sparse linear SVMs to generate linear models that can generalize well, and uses a subset of nonzero weighted variables found by the linear models to produce a final nonlinear model. The method exploits the fact that a linear SVM (no kernels) with l1-norm regularization inherently performs variable selection as a side-effect of minimizing capacity of the SVM model. The distribution of the linear model weights provides a mechanism for ranking and interpreting the effects of variables. Starplots are used to visualize the magnitude and variance of the weights for each variable. We illustrate the effectiveness of the methodology on synthetic data, benchmark problems, and challenging regression problems in drug design. This method can dramatically reduce the number of variables and outperforms SVMs trained using all attributes and using the attrib...

Background MicroRNAs (miRNAs), short ~21-nucleotide RNA molecules, play an important role in post-transcriptional regulation of gene expression. The number of known miRNA hairpins registered in the miRBase database is rapidly increasing, but recent reports suggest that many miRNAs with restricted temporal or tissue-specific expression remain undiscovered. Various strategies for in silico miRNA identification have been proposed to facilitate miRNA discovery. Notably support vector machine (SVM) methods have recently gained popularity. However, a drawback of these methods is that they do not provide insight into the biological properties of miRNA sequences. Results We here propose a new strategy for miRNA hairpin prediction in which the likelihood that a genomic hairpin is a true miRNA hairpin is evaluated based on statistical distributions of observed biological variation of properties (descriptors) of known miRNA hairpins. These distributions are transformed into a single and contin...

We present a data structure that stores a string $s[1..n]$ over the alphabet $[1..\sigma]$ in $nH_0(s) + o(n)(H_0(s){+}1)$ bits, where $H_0(s)$ is the zero-order entropy of $s$. % This data structure supports the operators \access\ and \rank\ in time $\Oh{\lg\lg\sigma}$, and the \select\ operator in constant time. % This result improves on previously known data structures using $nH_0(s)+o(n\lg\sigma)$ bits, where on highly compressible instances the redundancy $o(n\lg\sigma)$ can cease to be negligible compared to the $nH_0(s)$ bits that encode the data. % The technique is based on combining previous results through an ingenious partitioning of the alphabet, and practical enough to be implementable. It applies not only to strings, but also to several other compact data structures. % For example, we achieve $(i)$ the first representation of a string $s$ using $nH_k(s)+o(n)\lg\sigma$ bits and supporting \access, \rank, and \select\ in poly-loglog time; $(ii)$ faster search times for t...

Let π 1 ,. .. , π k be k (≥ 2) independent populations, where π i denotes the uniform distribution over the interval (0, θ i) and θ i > 0 (i = 1,. .. , k) is an unknown scale parameter. Let θ [1] ≤ · · · ≤ θ [k] be the ordered values of θ 1 ,. .. , θ k. The population π (k) (π (1)) associated with the unknown parameter θ [k] (θ [1]) is called the best (worst) population. For selecting the best population, we consider a general class of selection rules based on the natural estimators of θ i , i = 1,. .. , k. Under the entropy loss function, we consider the problem of estimating the scale parameter θ S of the population selected using a fixed selection rule from this class. We derive the uniformly minimum risk unbiased estimator of θ S and two natural estimators of θ S are also considered. We derive a general result for improving a scale invariant estimator of θ S under the entropy loss function. A simulation study on the performances of various competing estimators of θ S is also reported. Finally, we provide similar results for the problem of estimating the scale parameter of selected population when the selection goal is that of selecting the worst uniform population.

In this article, we present the problem of selecting a good stochastic system with high probability and minimum total simulation cost when the number of alternatives is very large. We propose a sequential approach that starts with the Ordinal Optimization procedure to ...