Hashing with Angular Reconstructive Embeddings.

Large-scale search methods are increasingly critical for many content-based visual analysis applications, among which hashing-based approximate nearest neighbor search techniques have attracted broad interests due to their high efficiency in storage and retrieval. However, existing hashing works are commonly designed for measuring data similarity by the Euclidean distances. In this paper, we focus on the problem of learning compact binary codes using the cosine similarity. Specifically, we proposed novel angular reconstructive embeddings (ARE) method, which aims at learning binary codes by minimizing the reconstruction error between the cosine similarities computed by original features and the resulting binary embeddings. Furthermore, we devise two efficient algorithms for optimizing our ARE in continuous and discrete manners, respectively. We extensively evaluate the proposed ARE on several large-scale image benchmarks. The results demonstrate that ARE outperforms several state-of-the-art methods.

Bilinear Optimized Product Quantization for Scalable Visual Content Analysis.

Product quantization (PQ) has been recognized as a useful technique to encode visual feature vectors into compact codes to reduce both the storage and computation cost. Recent advances in retrieval and vision tasks indicate that high-dimensional descriptors are critical to ensuring high accuracy on large-scale data sets. However, optimizing PQ codes with high-dimensional data is extremely time-consuming and memory-consuming. To solve this problem, in this paper, we present a novel PQ method based on bilinear projection, which can well exploit the natural data structure and reduce the computational complexity. Specifically, we learn a global bilinear projection for PQ, where we provide both non-parametric and parametric solutions. The non-parametric solution does not need any data distribution assumption. The parametric solution can avoid the problem of local optima caused by random initialization, and enjoys a theoretical error bound. Besides, we further extend this approach by learning locally bilinear projections to fit underlying data distributions. We show by extensive experiments that our proposed method, dubbed bilinear optimization product quantization, achieves competitive retrieval and classification accuracies while having significant lower time and space complexities.

Joint Hierarchical Category Structure Learning and Large-Scale Image Classification.

We investigate the scalable image classification problem with a large number of categories. Hierarchical visual data structures are helpful for improving the efficiency and performance of large-scale multi-class classification. We propose a novel image classification method based on learning hierarchical inter-class structures. Specifically, we first design a fast algorithm to compute the similarity metric between categories, based on which a visual tree is constructed by hierarchical spectral clustering. Using the learned visual tree, a test sample label is efficiently predicted by searching for the best path over the entire tree. The proposed method is extensively evaluated on the ILSVRC2010 and Caltech 256 benchmark datasets. The experimental results show that our method obtains significantly better category hierarchies than other state-of-the-art visual tree-based methods and, therefore, much more accurate classification.

A Fast Optimization Method for General Binary Code Learning.

Hashing or binary code learning has been recognized to accomplish efficient near neighbor search, and has thus attracted broad interests in recent retrieval, vision, and learning studies. One main challenge of learning to hash arises from the involvement of discrete variables in binary code optimization. While the widely used continuous relaxation may achieve high learning efficiency, the pursued codes are typically less effective due to accumulated quantization error. In this paper, we propose a novel binary code optimization method, dubbed discrete proximal linearized minimization (DPLM), which directly handles the discrete constraints during the learning process. Specifically, the discrete (thus nonsmooth nonconvex) problem is reformulated as minimizing the sum of a smooth loss term with a nonsmooth indicator function. The obtained problem is then efficiently solved by an iterative procedure with each iteration admitting an analytical discrete solution, which is thus shown to converge very fast. In addition, the proposed method supports a large family of empirical loss functions, which is particularly instantiated in this paper by both a supervised and an unsupervised hashing losses, together with the bits uncorrelation and balance constraints. In particular, the proposed DPLM with a supervised ℓ2 loss encodes the whole NUS-WIDE database into 64-b binary codes within 10 s on a standard desktop computer. The proposed approach is extensively evaluated on several large-scale data sets and the generated binary codes are shown to achieve very promising results on both retrieval and classification tasks.

Risk factors in women with gestational diabetes mellitus / 中华预防医学杂志

<p><b>OBJECTIVE</b>To identify risk factors for gestational diabetes mellitus (GDM) in women and to study the contribution of family history of type-2 diabetes to the risk for DDM.</p><p><b>METHODS</b>A case-control study was performed in 85 women with GDM and 177 cases controls. Univariate and multivariate logistic regression and log-linear model were used to identify risk factors of GDM.</p><p><b>RESULTS</b>Multivariate logistic regression showed that obesity before pregnancy, family history of type-2 diabetes, birth weight of pregnant women, age, fasting plasma level of triglyceride, physical inactivity, etc. all were risk factors for GDM. Analysis with log-linear model showed that parents' (father's or mother's) history of type-2 diabetes associated with GDM, with P-values of 0.012 and 0.017, respectively. Prevalence of diabetes in the mothers of proband with GDM was 9.41%, as compared with that in the fathers of proband with GDM 8.24%, with no statistical significance.</p><p><b>CONCLUSIONS</b>Obesity before gestation, family history of type-2 diabetes, low birth weight of mother, age, increased fasting plasma level of triglyceride, as well as parents' history of type-2 diabetes, all were risk factors for GDM. Physical exercise was found to be a protective factor for GDM. Mother's history of type-2 diabetes did not differ from father's in contributing to the onset of GDM in their offspring.</p>

Genetic polymorphism of UDP-glucuronosyltransferase 1F and susceptibility to hepatocellular carcinoma / 中华医学遗传学杂志

<p><b>OBJECTIVE</b>To study the genetic polymorphisms of UDP-glucuronosyltransferase 1F(UGT1F) and the relationship between polymorphisms and susceptibility to hepatocellular carcinoma (HCC).</p><p><b>METHODS</b>The polymorphisms of UGT1F of 84 patients with HCC and 144 healthy controls were detected by PCR-denaturation gradient gel electrophoresis-sequencing or PCR-single strain conformation polymorphsim-sequencing.</p><p><b>RESULTS</b>Three new single nucleotide polymorphisms(SNP) were found: the first one was a transversion of TrarrG at nucleotide 232; the second one was the transition of ArarrG at nucleotide 528 in exon 1; the last one was the transition of ArarrG at nucleotide 376 in intron 2. Additionally, the polymorphism at nucleotide 754 was proved in this study. The frequencies of genotype and allele of 4 loci in cases and controls were analyzed. Both frequencies of genotype G/G(13.10%) and allele G (29.17%) of position 754 of UGT1F in cases were sig nificantly greater than those in controls (2.78% and 19.44% ) respectively. For other loci, the difference between the two groups were not significant.</p><p><b>CONCLUSION</b>Exons 2-5 of UGT1F are highly conservative, but exon 1 emerges highly polymorphic. And the polymorphism at locus 754 may be related with HCC.</p>

A study on the genetic epidemiology of hepatocellular carcinoma / 中华流行病学杂志

<p><b>OBJECTIVE</b>To explore the interaction between inheritance and environment with the aid of research on the genetic modes of hepatocellular carcinoma (HCC).</p><p><b>METHODS</b>A genetic epidemiological study of HCC was conducted based on the methods of Penrose, simple segregation and Falconer for 100 proband pedigrees from HBsAg positive cohort. The proband samples came from a cohort of 90,00 people who were followed for 8 years. Analyses on genetic modes were carried out and heritability was calculated through the comparison of the proband pedigrees incidence frequency with incidence frequencies of the cohort and general population.</p><p><b>RESULTS</b>The incidence frequency of first-degree relatives was 4.0%, higher than what was seen in the general population incidence frequency (0.44%) and the cohort (1.03%). A familial aggregation of HBsAg carriers and a strong positive correlation between HBsAg carrier status and HCC were noticed (OR = 8.44, 95% CI: 3.37-20.06, P < 0.001). A ratio of the incidence frequency among siblings to the incident frequency among general population (s/q) approached 1/q(1/2) by Penrose method, but simple segregation did not show agreement with single-gene inheritance. The heritability from positive cohort was 42% +/- 6% (P < 0.05), compared with the heritability (59% +/- 7%) of general population. When the effect of the HBsAg was under control, the heritability from positive cohort turned to be 29% +/- 8% (P < 0.05), compared with the heritability (47% +/- 7%) of general population.</p><p><b>CONCLUSION</b>Our findings suggested that HCC followed a multifactorial mode rather than single inheritance. An interaction effect of inheritance and environment on HCC was also noticed.</p>

An association between gene polymorphism of tumor necrosis factor ? and systemic lupus erythematosus / 中华风湿病学杂志

Objective To make sure whether there is any association between genetic polymorphism of tumor necrosis factor (TNF) ? and systemic lupus erythematosus (SLE).Method PCR RFLP was used.A population based and family based study was carried out in 99 SLE patients,116 health controls and 12 families.Results The TNF ?2 allele frequency of SLE patients was significantly different from that of controls ( P

Mutations in BRCA1 and BRCA2 in sporadic cases with breast cancer in eastern China / 中国癌症杂志

Purpose:To estimate prevalence of BRCA1 and BRCA2 mutations among Chinese patients with breast cancer in eastern China. Methods:Frozen tumor tissues were obtained from 79 female patients, and three exons (2, 11, 22) of BRCA1 and three exons (9, 14, 22) of BRCA2 were examined with 23 pairs of primer, using polymerase chain reaction (PCR) single strand conformational polymorphism (SSCP) sequencing method. Results:It was found that in cDNA of BRCA1, there is a single nucleotide variation of T→C at position of 2430. By the RFLP method, we confirmed that the variation in the general population is a single nucleotide polymorphism. The distribution of T allele and C allele is different between cases and controls, but the difference is not significant. Conclusions:It implied that the mutations of BRCA1 and BRCA2 in breast cancer patients in eastern China is very rare.