Symmetry breaking and training from incomplete data with Radial Basis Boltzmann Machines

A Radial Basis Boltzmann Machine (RBBM) is a specialized Boltzmann Machine architecture that combines feed-forward mapping with probability estimation in the input space, and for which very efficient learning rules exist. The hidden representation of the network displays symmetry breaking as a function of the noise in the dynamics. Thus, generalization can be studied as a function of the noise in the neuron dynamics instead of as a function of the number of hidden units. We show that the RBBM can be seen as an elegant alternative of k-nearest neighbor, leading to comparable performance without the need to store all data. We show that the RBBM has good classification performance compared to the MLP. The main advantage of the RBBM is that simultaneously with the input-output mapping, a model of the input space is obtained which can be used for learning with missing values. We derive learning rules for the case of incomplete data, and show that they perform better on incomplete data than the traditional learning rules on a 'repaired' data set.     

Plasma and erythrocyte alpha-tocopherol and plasma retinol concentrations in term infants fed formula enriched with long-chain polyunsaturated fatty acids

OBJECTIVE: To assess the effect of long-chain polyunsaturated fatty acids (LCPUFA)- and vitamin E-supplemented formula feeding on erythrocyte and plasma alpha-tocopherol (VE), and plasma retinol (VA) concentrations in neonates and to compare these values with those found in infants feeding on infant formula without LCPUFA or breast milk SETTING: University Hospital of Granada, Spain. SUBJECTS: 49 full-term infants. DESIGN AND INTERVENTION: Subjects who chose not to breast feed were fed either (i) unsupplemented infant formula (F) or (ii) infant formula supplemented with LCPUFA and vitamin E (FL). Alpha-tocopherol and retinol were measured at 7 days, 1 month and 3 months. RESULTS: Plasma and erythrocyte VE concentrations and plasma VE/total lipids ratio increased significantly in all groups at 1 month of life (P < 0.05), but did not change significantly between 1 month and 3 months in any group (P > 0.05). Erythrocyte VE and VA retinol concentrations were higher in infants fed an infant formula than in breast milk-fed infants at 1 month of life (P < 0.05). Finally, there were no significant differences in plasma or erythrocyte VE levels, plasma VA or plasma VE/total lipid ratio between any groups at 3 months of life (P > 0.05). CONCLUSION: Infants fed on LCPUFA- and vitamin E-supplemented infant formula for 3 months have similar vitamin E and A status to infants fed on breast milk or infant formula without LCPUFA supplementation.     

Synthesis and comparative molecular field analysis (CoMFA) of antitumor 3-arylisoquinoline derivatives

In this study a series of 3-arylisoquinoline derivatives were synthesized and cytotoxicity against human melanoma tumor cell evaluated, and a three dimensional quantitative structure-activity relationship was investigated using the comparative molecular field analysis (CoMFA). The results suggested that the electrostatic, steric and hydrophobic factors of 3-arylisoquinolines were strongly correlated with the antitumor activity. Considerable predictive ability (cross-validated r2 as high as 0.721) was obtained through CoMFA.     

What is "special" about face perception?

There is growing evidence that face recognition is "special" but less certainty concerning the way in which it is special. The authors review and compare previous proposals and their own more recent hypothesis, that faces are recognized "holistically" (i.e., using relatively less part decomposition than other types of objects). This hypothesis, which can account for a variety of data from experiments on face memory, was tested with 4 new experiments on face perception. A selective attention paradigm and a masking paradigm were used to compare the perception of faces with the perception of inverted faces, words, and houses. Evidence was found of relatively less part-based shape representation for faces. The literatures on machine vision and single unit recording in monkey temporal cortex are also reviewed for converging evidence on face representation. The neuropsychological literature is reviewed for-evidence on the question of whether face representation differs in degree or kind from the representation of other types of objects.