Very small sets

Towards Optimal Learning from Very Small Data Sets

The problem of generalization in neural networks, that is, how well a network will perform on unseen data, is has received much attention recently. In this paper we present an approach to generating learning algorithms which have the potential to generalize from very small training sets. We believe that in this paper outlines a new, although potentially computationally expensive, approach to op...

The Very Small Array

The Very Small Array (VSA) is a fourteen-element interferometer designed to study the cosmic microwave background on angular scales of 2.4 to 0.2 degrees (angular multipoles l = 150 to 1800). It operates at frequencies between 26 and 36 GHz, with a bandwidth of 1.5 GHz, and is situated at the Teide Observatory, Tenerife. The instrument also incorporates a single-baseline interferometer, with la...

Very small P values.

A Short Note on Using Multiple Imputation Techniques for Very Small Data Sets

This short note describes a simple experiment to investigate the value of using multiple imputation (MI) methods [2, 3]. We are particularly interested in whether a simple bootstrap based on a k-nearest neighbour (kNN) method can help address the problem of missing values in two very small, but typical, software project data sets. This is an important question because, unfortunately, many real-...

Small and Very Small Interstellar Grains

This review summarizes the observational characteristics of those interstellar grains, which are prevented from entering the solar system by interactions with the heliopause. Such grains are typically less than 100 nm in radius and they reveal their presence by interstellar absorption at ultraviolet wavelengths and by non-equilibrium emissions in the red and near-infrared portions of the spectrum.