Supplementary data are available at Bioinformatics online.

No abstract

No abstract

The order Archaeognatha was an ancient group of Hexapoda and was considered as the most primitive of living insects. Two extant families (Meinertellidae and Machilidae) consisted of approximately 500 species. This study determined 3 complete mitochondrial genomes and 2 nearly complete mitochondrial genome sequences of the bristletail. The size of the 5 mitochondrial genome sequences of bristletail were relatively modest, containing 13 protein-coding genes (PCGs), 2 ribosomal RNA (rRNA) genes, 22 transfer RNA (tRNA) genes and one control region. The gene orders were identical to that of Drosophila yakuba and most bristletail species suggesting a conserved genome evolution within the Archaeognatha. In order to estimate archaeognathan evolutionary relationships, phylogenetic analyses were conducted using concatenated nucleotide sequences of 13 protein-coding genes, with four different computational algorithms (NJ, MP, ML and BI). Based on the results, the monophyly of the family Machilidae was challenged by both datasets (W12 and G12 datasets). The relationships among archaeognathan subfamilies seemed to be tangled and the subfamily Machilinae was also believed to be a paraphyletic group in our study.

Abstract-Wireless mesh networks are widely applied in many fields such as industrial controlling, environmental monitoring and militar y operations. Network coding is promising technology that can improve the performance of wireless mesh networks. In particular, network coding is suitable for wireless mesh networks as the fixed backbone of wireless mesh is usually unlimited energy. However, coding collision is a severe problem affecting network performance. To avoid this, routing should be effectively designed with an optimum combination of coding opportunity and coding validity. In this paper, we propose a Connected Dominating Set (CDS)-based and Flow-oriented Coding-aware Routing (CFCR) mechanism to actively increase potential coding opportunities. Our work provides two major contributions. First, it effectively deals with the coding collision problem of flows by introducing the information conformation process, which effectively decreases the failure rate of decoding. Secondly, our routing process considers the benefit of CDS and flow coding simultaneously. Through formalized analysis of the routing parameters, CFCR can choose optimized routing with reliable transmission and small cost. Our evaluation shows CFCR has a lower packet loss ratio and higher throughput than existing methods, such as Adaptive Control of Packet Overhead in XOR Network Coding (ACPO), or Distributed Coding-Aware Routing (DCAR).

By injecting adversarial examples into training data, adversarial training ispromising for improving the robustness of deep learning models. However, most existing adversarial training approaches are based on a specific type of adversarial attack. It may not provide sufficiently representative samples from the adversarial domain, leading to a weak generalization ability on adversarial examples from other attacks. Moreover, during the adversarial training, adversarial perturbations on inputs are usually crafted by fast single-step adversaries so as to scale to large datasets. This work is mainly focused on the adversarial training yet efficient FGSM adversary. In this scenario, it is difficult to train a model with great generalization due to the lack of representative adversarial samples, aka the samples are unable to accurately reflect the adversarial domain. To alleviate this problem, we propose a novel Adversarial Training with Domain Adaptation (ATDA) method. Our intuition is to regard the adversarial training on FGSM adversary as a domain adaption task with limited number of target domain samples. The main idea is to learn a representation that is semantically meaningful and domain invariant on the clean domain as well as the adversarial domain. Empirical evaluations on Fashion-MNIST, SVHN, CIFAR-10 and CIFAR-100 demonstrate that ATDA can greatly improve the generalization of adversarial training and the smoothness of the learned models, and outperforms state-of-the-art methods on standard benchmark datasets. To show the transfer ability of our method, we also extend ATDA to the adversarial training on iterative attacks such as PGD-Adversial Training (PAT) and the defense performance is improved considerably.

No abstract

The main purpose of this paper is to provide an analytical proof of the inadmissibility of the classical confidence interval for the component of the mean of a multivariate normal distribution in terms of coverage probability and interval length in the sense of "parametric empirical Bayes theory" as defined in Morris (1983a, b). We adapt the technique of Casella and Hwang (1983), centered at the positive part of the James-Stein estimator with variable radius, to construct a new confidence interval which is shown to be nearly better than the classical one numerically when dimension p > 3. Also, some comparisons with the naive interval, Morris interval (1983a,b) and bootstrap interval (Laird and Louis, 1987) are performed.