Regularization and complexity control in feed-forward networks

In this paper we consider four alternative approaches to complexity control in feed-forward networks based respectively on architecture selection, regularization, early stopping, and training with noise. We show that there are close similarities between these approaches and we argue that, for most practical applications, the technique of regularization should be the method of choice.

The relationship between parental perceptions of diabetes and glycaemic control

Aim: To measure the relationship between perceived child competence, parental self-efficacy, and children's glycaemic control. Methods: Cross-sectional outpatient based questionnaire survey of 78 parents of children aged 6-12 years with insulin dependent diabetes mellitus, diagnosed for at least one year. Parental perceptions of their child's competence were assessed, together with parental perceptions of their own self-efficacy in managing their child's diabetes. Glycaemic control was assessed by the average annual HbA 1C level. Results: The response rate was 64.5% (51 parents); 82% were mothers and the socioeconomic class and ethnicity spread was representative of the general population. The mean age of the children was 10 years and duration of diabetes 4.4 years. Poorer glycaemic control was associated with higher perceived child competence, together with lower perceived age of responsibility, lower perceived seriousness, and less frequent blood tests. Higher parental self-efficacy and higher perceived child competence predicted a higher level of normalisation, as did lower perceived seriousness, a lower perceived parental responsibility for management, and a less protective style of parenting. Conclusion: Parents' perceptions of their children's diabetes are significantly related to glycaemic control; however, those who appear more competent at managing diabetes may overestimate their child's capabilities, leading to poorer glycaemic control.