Deep neural nets with a large number of parameters are very powerful machine learning systems. However, overfitting is a serious problem in such networks. Large networks are also slow to use, making it difficult to deal with overfitting by combining the predictions of many different large neural nets at test time. Dropout is a technique for addressing this problem. The key idea is to randomly drop units (along with their connections) from the neural network during training. This prevents units from co-adapting too much. During training, dropout samples from an exponential number of different "thinned" networks. At test time, it is easy to approximate the effect of averaging the predictions of all these thinned networks by simply using a single unthinned network that has smaller weights. This significantly reduces overfitting and gives major improvements over other regularization methods. We show that dropout improves the performance of neural networks on supervised learning tasks in vision, speech recognition, document classification and computational biology, obtaining state-of-the-art results on many benchmark data sets.

/pdf/dropout-a-simple-way-to-prevent-neural-networks-from-4gtr92okhv.pdf

Dropout: a simple way to prevent neural networks from overfitting

抗原变异可使得多种致病微生物易于逃避宿主免疫应答。表达在感染红细胞表面的恶性疟原虫红细胞表面蛋白1（PfPMP1）与感染红细胞、内皮细胞、树突状细胞以及胎盘的单个或多个受体作用，在黏附及免疫逃避中起关键的作用。每个单倍体基因组var基因家族编码约60种成员，通过启动转录不同的var基因变异体为抗原变异提供了分子基础。

疟原虫var基因转换速率变化导致抗原变异[英]／Paul H, Robert P, Christodoulou Z, et al//Proc Natl Acad Sci U S A

The identification of genetically homogeneous groups of individuals is a long standing issue in population genetics. A recent Bayesian algorithm implemented in the software STRUCTURE allows the identification of such groups. However, the ability of this algorithm to detect the true number of clusters (K) in a sample of individuals when patterns of dispersal among populations are not homogeneous has not been tested. The goal of this study is to carry out such tests, using various dispersal scenarios from data generated with an individual-based model. We found that in most cases the estimated 'log probability of data' does not provide a correct estimation of the number of clusters, K. However, using an ad hoc statistic DeltaK based on the rate of change in the log probability of data between successive K values, we found that STRUCTURE accurately detects the uppermost hierarchical level of structure for the scenarios we tested. As might be expected, the results are sensitive to the type of genetic marker used (AFLP vs. microsatellite), the number of loci scored, the number of populations sampled, and the number of individuals typed in each sample.

/pdf/detecting-the-number-of-clusters-of-individuals-using-the-16mlx2qlv8.pdf

Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study.

Arlequin ver 3.0 is a software package integrating several basic and advanced methods for population genetics data analysis, like the computation of standard genetic diversity indices, the estimation of allele and haplotype frequencies, tests of departure from linkage equilibrium, departure from selective neutrality and demographic equilibrium, estimation or parameters from past population expansions, and thorough analyses of population subdivision under the AMOVA framework. Arlequin 3 introduces a completely new graphical interface written in C++, a more robust semantic analysis of input files, and two new methods: a Bayesian estimation of gametic phase from multi-locus genotypes, and an estimation of the parameters of an instantaneous spatial expansion from DNA sequence polymorphism. Arlequin can handle several data types like DNA sequences, microsatellite data, or standard multi-locus genotypes. A Windows version of the software is freely available on http://cmpg.unibe.ch/software/arlequin3.

/pdf/arlequin-version-3-0-an-integrated-software-package-for-4cs6nd3y2p.pdf

Arlequin (version 3.0): An integrated software package for population genetics data analysis

We present here a new version of the Arlequin program available under three different forms: a Windows graphical version (Winarl35), a console version of Arlequin (arlecore), and a specific console version to compute summary statistics (arlsumstat). The command-line versions run under both Linux and Windows. The main innovations of the new version include enhanced outputs in XML format, the possibility to embed graphics displaying computation results directly into output files, and the implementation of a new method to detect loci under selection from genome scans. Command-line versions are designed to handle large series of files, and arlsumstat can be used to generate summary statistics from simulated data sets within an Approximate Bayesian Computation framework.

Arlequin suite ver 3.5: a new series of programs to perform population genetics analyses under Linux and Windows

In this chapter, we use data from a survey of rural residents in Xi’an, Yan’an and other places undertaken by Xi’an Jiaotong University to perform an empirical analysis of the impact of ecological compensation policies on rural livelihoods and on the poverty alleviation efforts in the western mountainous areas.

An Empirical Study of the Impact of Ecological Compensation Policy on Rural Households in the Western Mountainous Regions

Some topics in theoretical population genetics.

In this paper, we examine the conditions that allow increased recombination to evolve in the presence of recurrent deleterious mutation. We focus on a three-locus model first studied by Feldman et al. (1980), which follows the dynamics of a modifier locus that alters the recombination rate between two loci subject to deleterious mutation. Although Feldman et al. (1980) indicated that increased recombination might be favored if there is diminishing-returns epistasis between the loci under selection. Even with synergistic epistasis, evolution at the modifier locus will lead to decreased recombination if the modifier locus is loosely linked and epistasis is strong. Using multi-locus analysis of Barton (1995), we show that variability among loci in the sign and strength of epistasis also tends to select for decreased recombination. We conclude that the conditions that favor increased recombination in the presence of deleterious mutations are fairly restrictive, especially if, as seems likely, epistatic interactions are highly variable among loci.

Marcus W. Feldman

Papers

An Empirical Study of the Impact of Ecological Compensation Policy on Rural Households in the Western Mountainous Regions

Some topics in theoretical population genetics.

Deleteriuous Mutations, Variable Epistatic Interactions, and the Evolution of Recombination

2. The Evidence for Niche Construction

Conformity and anti-conformity in a finite population.