An Information Theoretic Interpretation to Deep Neural Networks.

With the unprecedented performance achieved by deep learning, it is commonly believed that deep neural networks (DNNs) attempt to extract informative features for learning tasks. To formalize this intuition, we apply the local information geometric analysis and establish an information-theoretic framework for feature selection, which demonstrates the information-theoretic optimality of DNN features. Moreover, we conduct a quantitative analysis to characterize the impact of network structure on the feature extraction process of DNNs. Our investigation naturally leads to a performance metric for evaluating the effectiveness of extracted features, called the H-score, which illustrates the connection between the practical training process of DNNs and the information-theoretic framework. Finally, we validate our theoretical results by experimental designs on synthesized data and the ImageNet dataset.

Modulation of the Gut Microbiota and Liver Transcriptome by Red Yeast Rice and Monascus Pigment Fermented by Purple Monascus SHM1105 in Rats Fed with a High-Fat Diet.

Hyperlipidemia can easily cause atherosclerosis and induce cardiovascular and cerebrovascular diseases. Red yeast rice (RYR) contains a variety of active ingredients and is commonly used as medicine and food, and has pharmacological effects such as lowering blood lipids. In this study, we select Monascus strain SHM1105 with a high yield of Monacolin K and monascus pigment (PIG), and studied the effects of the RYR and PIG fermented by this strain on blood lipids, intestinal flora, and liver transcriptome in hyperlipidemia model rats. The experimental results show that, compared with the high-fat model group, the weight growth rate, liver weight ratio, kidney weight ratio, spleen weight ratio, and fat weight ratio of rats in the gavage lovastatin (LOV), RYR, and PIG group were all significantly decreased (p < 0.05). Intervention with RYR and PIG can significantly reduce the serum TC, TG, and LDL-C levels, which has the effect of lowering blood lipids. The 16SrDNA sequencing results showed that the ratio of Firmicutes/Bacteroidetes decreased significantly (p ≤ 0.01) after the intervention of LOV, RYR, and PIG; the abundance of the ratio of Lachnospiraceae, Ruminococcaceae, Prevotellaceae, and Bacteroidales-S24-7-group also changed. The combined analysis of transcriptome and metabolome showed that lovastatin, RYR, and PIG can all improve lipid metabolism in rats by regulating Steroid hormone biosynthesis, Glycerolipid metabolism, and the Arachidonic acid metabolism pathway. In addition, RYR and PIG also have a unique way of regulating blood lipids. Although a lot of research on the lipid-lowering components of Monascus rice and the single pigment component of Monascus has been carried out, the actual application is RYR and pigments as mixtures, as a mixture of RYR and PIG contains a variety of biologically active ingredients, and each component may have a synergistic effect. Hence it has a lipid-lowering mechanism that lovastatin does not have. Therefore, RYR and PIG are effective in reducing lipid potential development and can be utilized in functional foods.