We present a parameter free and monotonic alternative to the parametric variable precision model of rough set data analysis, based on the well known PRE index λ of Goodman and Kruskal. Using weighted (parametric) λ models we show how expert knowledge can be integrated without losing the monotonic property of the index. Based on a weighted λ index we present a polynomial algorithm to determine a...

The problem of evaluating the solvency of insurance companies is tackled through the use of parametric and non-parametric statistical models, constructed using respectively multivariate methodologies and decision-tree techniques. First we intend to present the theoretical framework of both methodologies describing them within a Bayesian context, then two models are tested on a sample of Italian...

Online learning is discussed from the viewpoint of Bayesian statistical inference. By replacing the true posterior distribution with a simpler parametric distribution, one can define an online algorithm by a repetition of two steps: An update of the approximate posterior, when a new example arrives, and an optimal projection into the parametric family. Choosing this family to be Gaussian, we sh...

The explosive growth in data volume and the availability of cheap computing resources have sparked increasing interest in Big learning, an emerging subfield that studies scalable machine learning algorithms, systems and applications with Big Data. Bayesian methods represent one important class of statistical methods for machine learning, with substantial recent developments on adaptive, flexibl...

Phase I trials aim to establish appropriate clinical and statistical parameters to guide future clinical trials. With individual trials typically underpowered, systematic reviews and meta-analysis are desired to assess the totality of evidence. A high percentage of zero or missing outcomes often complicate such efforts. We use a systematic review of pediatric phase I oncology trials as an examp...

Health economic decision models compare costs and health effects of different interventions over the long term and usually incorporate survival data. Since survival is often extrapolated beyond the range of the data, inaccurate model specification can result in very different policy decisions. However, in this area, flexible survival models are rarely considered, and model uncertainty is rarely...

Multi-state models provide a unified framework for the description of the evolution of discrete phenomena in continuous time. One particular example are Markov processes which can be characterised by a set of time-constant transition intensities between the states. In this paper, we will extend such parametric approaches to semiparametric models with flexible transition intensities based on Bay...

Selection models are appropriate when the probability that a potential datum enters the sample is a nondecreasing function of the numeric value of the datum. It is rarely justiiable to model this function, called the weight function, with a speciic parametric form, but is appealing to model with a nonparametric prior centered around a parametric form. The Bayesian analysis with Dirichlet proces...

The objective of my doctoral research is bring together two fields: partially-observable reinforcement learning (PORL) and non-parametric Bayesian statistics (NPB) to address issues of statistical modeling and decisionmaking in complex, realworld domains.

The objective of my doctoral research is bring together two fields: partially-observable reinforcement learning (PORL) and non-parametric Bayesian statistics (NPB) to address issues of statistical modeling and decisionmaking in complex, realworld domains.