No free lunch for cross validation

It is known theoretically that an algorithm cannot be good for an arbitrary prior. We show that in practical terms this also applies to the technique of ``cross validation'', which has been widely regarded as defying this general rule. Numerical examples are analysed in detail. Their implications to researches on learning algorithms are discussed.

General gaussian priors for improved generalisation

We explore the dependence of performance measures, such as the generalization error and generalization consistency, on the structure and the parameterization of the prior on `rules', instanced here by the noisy linear perceptron. Using a statistical mechanics framework, we show how one may assign values to the parameters of a model for a `rule' on the basis of data instancing the rule. Information about the data, such as input distribution, noise distribution and other `rule' characteristics may be embedded in the form of general gaussian priors for improving net performance. We examine explicitly two types of general gaussian priors which are useful in some simple cases. We calculate the optimal values for the parameters of these priors and show their effect in modifying the most probable, MAP, values for the rules.

Survey of General Optical Council disciplinary and fitness to practise hearings:2001-2011

The aim of this survey was to review 187 transcripts from the United Kingdom’s General Optical Council (GOC) Disciplinary and Fitness To Practise (FTP) Committee hearings from 2001 to 2011 in order to identify common themes and thereby help practitioners to avoid the more frequently occurring pitfalls that were recorded during this period. The study covered changes in GOC FTP regulations in 2005, which involved a change from a disciplinary to a fitness to practise process. The number of cases was very small compared to the total number of optometrist and dispensing optician registrants, which was 13709 in 2001-02 rising to 18582 in 2010-11. The main findings indicated that between 2001 and 2011 there was a three times greater likelihood that male registrants versus female registrants would be brought in front of a GOC Disciplinary or FTP Committee. In terms of erasures from the GOC registers between 2001 and 2011, male registrants were also more likely to be erased than females. The male: female split for erasures between 2001 and 2011 was five: one, increasing to seven: one when considering the situation post the 2005 GOC FTP rule change. Of the cases brought before the Disciplinary and FTP Committees between 2001 and 2011, it was noted that cases implicating theft and fraud were most frequent representing 27% of hearings examined (17% involving NHS fraud and 10% theft or fraud from an employer). The examination of transcripts revealed other hearings were more complex. These hearings often had a primary reason for the investigation that highlighted further secondary concerns that also required investigation.