Entropy-Randomized Method for the Reconstruction of Missing Data

Automation and Remote Control - The article deals with the problem of reconstructing missing data in data collections for machine learning problems. We propose a new randomized method for missing...     

Spatial analysis of air pollution and cancer incidence rates in Haifa Bay, Israel

The Israel National Cancer Registry reported in 2001 that cancer incidence rates in the Haifa area are roughly 20% above the national average. Since Haifa has been the major industrial center in Israel since 1930, concern has been raised that the elevated cancer rates may be associated with historically high air pollution levels. This work tests whether persistent spatial patterns of metrics of chronic exposure to air pollutants are associated with the observed patterns of cancer incidence rates. Risk metrics of chronic exposure to PM₁₀, emitted both by industry and traffic, and to SO₂, a marker of industrial emissions, was developed. Ward-based maps of standardized incidence rates of three prevalent cancers: Non-Hodgkin's lymphoma, lung cancer and bladder cancer were also produced. Global clustering tests were employed to filter out those cancers that show sufficiently random spatial distribution to have a nil probability of being related to the spatial non-random risk maps. A Bayesian method was employed to assess possible associations between the morbidity and risk patterns, accounting for the ward-based socioeconomic status ranking. Lung cancer in males and bladder cancer in both genders showed non-random spatial patterns. No significant associations between the SO₂-based risk maps and any of the cancers were found. Lung cancer in males was found to be associated with PM₁₀, with the relative risk associated with an increase of 1 μg/m³ of PM₁₀ being 12%. Special consideration of wards with expected rates < 1 improved the results by decreasing the variance of the spatially correlated residual log-relative risk.     

Entropy-robust randomized forecasting under small sets of retrospective data

This paper suggests a new randomized forecasting method based on entropy-robust estimation for the probability density functions (PDFs) of random parameters in dynamic models described by the systems of linear ordinary differential equations. The structure of the PDFs of the parameters and measurement noises with the maximal entropy is studied. We generate the sequence of random vectors with the entropy-optimal PDFs using a modification of the Ulam–von Neumann method. The developed method of randomized forecasting is applied to the world population forecasting problem.     

A New Nonparametric Pairwise Clustering Algorithm Based on Iterative Estimation of Distance Profiles

We present a novel pairwise clustering method. Given a proximity matrix of pairwise relations (i.e. pairwise similarity or dissimilarity estimates) between data points, our algorithm extracts the two most prominent clusters in the data set. The algorithm, which is completely nonparametric, iteratively employs a two-step transformation on the proximity matrix. The first step of the transformation represents each point by its relation to all other data points, and the second step re-estimates the pairwise distances using a statistically motivated proximity measure on these representations. Using this transformation, the algorithm iteratively partitions the data points, until it finally converges to two clusters. Although the algorithm is simple and intuitive, it generates a complex dynamics of the proximity matrices. Based on this bipartition procedure we devise a hierarchical clustering algorithm, which employs the basic bipartition algorithm in a straightforward divisive manner. The hierarchical clustering algorithm copes with the model validation problem using a general cross-validation approach, which may be combined with various hierarchical clustering methods.We further present an experimental study of this algorithm. We examine some of the algorithm's properties and performance on some synthetic and standard data sets. The experiments demonstrate the robustness of the algorithm and indicate that it generates a good clustering partition even when the data is noisy or corrupted.     

Using Factor Oracles for Machine Improvisation

We describe variable markov models we have used for statistical learning of musical sequences, then we present the factor oracle, a data structure proposed by Crochemore & al for string matching. We show the relation between this structure and the previous models and indicate how it can be adapted for learning musical sequences and generating improvisations in a real-time context.