752 publications from this institution
Deep learning techniques have been paramount in the last years, mainly due to their outstanding results in a number of applications. In this paper, we address the issue of fine-tuning parameters of Deep Belief Networks by means of meta-heuristics in which real-valued decision variables are described by quaternions. Such approaches essentially perform optimization in fitness landscapes that are mapped to a different representation based on hypercomplex numbers that may generate smoother surfaces. We therefore can map the optimization process onto a new space representation that is more suitable to learning parameters. Also, we proposed two approaches based on Harmony Search and quaternions that outperform the state-of-the-art results obtained so far in three public datasets for the reconstruction of binary images.
Low latency is an important design goal for reliable data transmission protocols such as TCP and QUIC. However, timeout-based loss recovery can unnecessarily increase end-to-end latency. Previous work in reducing timeout-based loss recovery latency either duplicates every packet to avoid loss or focuses on fine-tuning the timeout timers to shorten the timeout latency without causing spurious packet retransmissions. In this work, we propose a new mechanism called Selective Loss Prevention (SLP) to reduce the loss recovery latency of a reliable transport protocol. Through extensive trace analysis, we find that not all lost packets are equal. The loss of packets with certain flags, such as SYN and PSH, is more likely to cause timeouts than other packets. Based on this observation, we propose to selectively duplicate an "important" packet whose loss is likely to increase a connection's latency. We design an algorithm to determine when to duplicate a lost packet proactively and incorporate it into TCP's congestion control algorithm so that duplicate packets will not congest the network. We incorporate SLP into Linux's kernel and evaluate its performance. Our results show that SLP can reduce timeout-based latency caused by the loss of important packets in a connection, and its overhead is low.
Conserved polycomb repressive complex 2 (PRC2) mediates H3K27me3 to direct transcriptional repression and has a key role in cell fate determination and cell differentiation in both animals and plants. PRC2 subunits have undergone independent multiplication and functional divergence in higher plants. However, relevant information is still absent in gymnosperms.To launch gymnosperm PRC2 research, we identified and cloned the PRC2 core component genes in the conifer model species Picea abies, including one Esc/FIE homolog PaFIE, two p55/MSI homologs PaMSI1a and PaMSI1b, two E(z) homologs PaKMT6A2 and PaKMT6A4, a Su(z)12 homolog PaEMF2 and a PaEMF2-like fragment. Phylogenetic and protein domain analyses were conducted. The Esc/FIE homologs were highly conserved in the land plant, except the monocots. The other gymnospermous PRC2 subunits underwent independent evolution with angiospermous species to different extents. The relative transcript levels of these genes were measured in endosperm and zygotic and somatic embryos at different developmental stages. The obtained results proposed the involvement of PaMSI1b and PaKMT6A4 in embryogenesis and PaKMT6A2 and PaEMF2 in the transition from embryos to seedlings. The PaEMF2-like fragment was predominantly expressed in the endosperm but not in the embryo. In addition, immunohistochemistry assay showed that H3K27me3 deposits were generally enriched at meristem regions during seed development in P. abies.This study reports the first characterization of the PRC2 core component genes in the coniferous species P. abies. Our work may enable a deeper understanding of the cell reprogramming process during seed and embryo development and may guide further research on embryonic potential and development in conifers.
Simulated annealing (SA) is a trajectory-based, random search technique for global optimization. It mimics the annealing process in materials processing when a metal cools and freezes into a crystalline state with minimum energy and larger crystal sizes so as to reduce the defects in metallic structures. The annealing process involves the careful control of temperature and its cooling schedule. SA has been successfully applied in many areas.
No abstract is provided for this article.
Mathematical optimization or mathematical programming consists of two main categories: linear programming and nonlinear programming. In this chapter, we will first introduce linear programming briefly.
Efficiency of an optimisation process is largely determined by the search algorithm and its fundamental characteristics. In a given optimisation, a single type of algorithm is used in most applications. In this paper, we will investigate the eagle strategy recently developed for global optimisation, which uses a two-stage strategy by combing two different algorithms to improve the overall search efficiency. We will discuss this strategy with differential evolution and then evaluate their performance by solving real-world optimisation problems such as pressure vessel and speed reducer design. Results suggest that we can reduce the computing effort by a factor of up to ten in many applications.
This chapter introduces some of the most widely used techniques for data mining, including nearest-neighbor algorithm, k-mean algorithm, decision trees, random forests, Bayesian classifier, and others. Special techniques such as CURE and BFR for mining big data are also briefly introduced.
The increasing popularity of metaheuristic algorithms has attracted a great deal of attention in algorithm analysis and performance evaluations. No-free-lunch theorems are of both theoretical and practical importance, while many important studies on convergence analysis of various metaheuristic algorithms have proven to be fruitful. This paper discusses the recent results on no-free-lunch theorems and algorithm convergence, as well as their important implications for algorithm development in practice. Free lunches may exist for certain types of problem. In addition, we will highlight some open problems for further research.
The aim of this article is to demonstrate that a firm may owe its continued existence to its attempts to conceal information from its competitors about the unknown characteristics of a certain factor, not just to its savings on market transaction costs, its team-working, risk-sharing or the encouragement of ex ante specific investment. This is because the existence of a firm contract severs the relationship between the factor market and the product market, thereby making it difficult for outsiders to observe the marginal contribution of the intermediate factor and make statistical inferences about the factor’s unknown characteristics. Furthermore, an optimal contract is determined by a trade-off not only between traditional risk-sharing and incentive, but also between the incentive and information concealing. Finally, we show that this latter kind of trade-off also affects the position of the optimal boundary of the firm.
No abstract is provided for this article.
No abstract is provided for this article.
Thyroid cancer (TC) is the most common malignancy of the endocrine system and its incidence is gradually rising. Research has demonstrated a close link between autophagy and thyroid cancer. We constructed a prognostic model of autophagy-related long noncoding RNA (lncRNA) in thyroid cancer and explored its prognostic value. A total of 14,142 lncRNAs and 212 autophagy-related genes (ATGs) were obtained from the Cancer Genome Atlas (TCGA) database and the Human Autophagy Database (HADb), respectively. We performed lncRNA-ATGs correlation analysis and finally obtained 1166 autophagy-associated lncRNAs. Subsequently we conducted univariate Cox regression analysis and multivariate Cox regression analysis, a nine-autophagy-related lncRNAs (AC092279.1, AC096677.1, DOCK9-DT, LINC02454, AL136366.1, AC008063.1, AC004918.3, LINC02471, AL162231.2) significantly associated with prognosis was identified. Based on these autophagy-related lncRNAs, a risk model was constructed. The area under the curve (AUC) of the risk score was 0.905, proving that the accuracy of risk signature was superior. In addition, multiple regression analysis showed that risk score was a significant independent prognostic risk factor for thyroid cancer. In this study, a nine autophagy-related lncRNAs in thyroid cancer were established to predict the prognosis of thyroid cancer patients.
Flower pollination algorithm is a new nature-inspired algorithm, based on the characteristics of flowering plants. In this paper, we extend this flower algorithm to solve multi-objective optimization problems in engineering. By using the weighted sum method with random weights, we show that the proposed multi-objective flower algorithm can accurately find the Pareto fronts for a set of test functions. We then solve a bi-objective disc brake design problem, which indeed converges quickly.
