Deep learning is a popular machine technique and has been applied to many real-world problems, ranging from computer vision natural language processing. However, training deep neural network very time-consuming, especially on big data. It become difficult for single train large model over datasets. A solution distribute parallelize the process across multiple machines using parameter server framework. In this paper, we present distributed paradigm framework called Dynamic Stale Synchronous...

Selecting the right drugs for patients is a primary goal of precision medicine. In this article, we consider problem cancer drug selection in learning-to-rank framework. We have formulated as to accurately predicting 1) ranking positions sensitive and 2) orders among cell lines based on their responses drugs. developed new method, denoted pLETORg, that predicts structures each line via using latent vectors vectors. The pLETORg method learns such through explicitly enforcing that, list line,...

The bulk synchronous parallel (BSP) is a celebrated synchronization model for general-purpose computing that has successfully been employed distributed training of machine learning models. A prevalent shortcoming the BSP it requires workers to wait straggler at every iteration. To ameliorate this classic BSP, we propose ELASTICBSP aims relax its strict requirement. proposed offers more flexibility and adaptability during phase, without sacrificing on accuracy trained model. We also an...

Many real world networks are very large and constantly change over time. These dynamic exist in various domains such as social networks, traffic biological interactions. To handle downstream applications link prediction anomaly detection, it is essential for to be transferred into a low dimensional space. Recently, network embedding, technique that converts graph low-dimensional representation, has become increasingly popular due its strength preserving the structure of network. Efficient...

Abstract In view of the problem that conventional tracker does not adapt to abrupt motion, we propose a tracking algorithm based on hybrid extended ant lion optimizer with sine cosine (EALO-SCA) in this paper. Firstly, multiple elites is used replace single elite standard (ALO). The ALO (EALO) can enhance global exploration ability, which handle motion. Secondly, considering (SCA) has strong local exploitation operator, EALO-SCA proposed using advantages both EALO and SCA. approach improve...

Conversational AI has become an increasingly prominent and practical application of machine learning. How-ever, existing conversational techniques still suffer from var-ious limitations. One such limitation is a lack well-developed methods for incorporating auxiliary information that could help model understand context better. In this paper, we explore how persona-based improve the quality response generation in conversations. First, provide literature review focusing on current...

Compared to modelling observable data, it is more difficult choose a suitable distribution describe latent variables since no prior knowledge or information can be used and only normal nonparametric distributions are mainly applied random effects for generalized linear mixed models (GLMMs) in the literature. To enhance toolkit, this article investigates class of parametric skew elliptical multilevel binomial regression using Bayesian approach; includes normal, Students’ t-distributions...

In this paper, a novel approach to the fusion and recognition of face iris image based on wavelet features Kernel Fisher Discriminant Analysis (KFDA) is developed. Firstly, dimension reduced, noise eliminated, storage space saved efficiency improved by Discrete Wavelet Transform (DWT) image. Secondly, are extracted KFDA. Finally, Nearest Neighbor classifier selected perform recognition. Experimental results ORL database CASIA show that not only 'small sample problem' overcome KFDA, but also...

In the field of computer vision, data augmentation is widely used to enrich feature complexity training datasets with deep learning techniques. However, regarding generalization capabilities models, difference in artificial features generated by and natural visual has not been fully revealed. This study focuses on representation variable 'illumination', simulating its distribution degradation examining how techniques enhance model performance a classification task. Our goal investigate...

The amplitude of low-field nuclear magnetic resonance (NMR) is weak, and the echo buried in noise. reduction noise critical to accurately extract amplitude. Phase correction-adaptive line enhancement (PC-ALE) proposed suppression based on principle ALE NMR spin-echo characteristics. calculated after two-stage processes; phase shift from time-delay filter tap would be compensated effectively frequency domain. Simulation experiments show that PC-ALE has prominent performance suppression,...

Effective in silico compound prioritization is a critical step to identify promising drug candidates the early stages of discovery. Current computational methods for usually focus on ranking compounds based one property, typically activity, with respect single target. However, selectivity also key property which should be deliberated simultaneously so as minimize likelihood undesired side effects future drugs. In this paper, we present novel machine-learning differential method dCPPP. This...

Humans use natural language to compose common concepts from their environment into plausible, day-to-day scene descriptions. However, such generative commonsense reasoning (GCSR) skills are lacking in state-of-the-art text generation methods. Descriptive sentences about arbitrary generated by neural models (e.g., pre-trained text-to-text Transformers) often grammatically fluent but may not correspond human sense, largely due lack of mechanisms capture concept relations, identify implicit...

Combined with diagonal image transform, two-dimensional discrete cosine transform (2DDCT) is used in face and iris for feature compression; then Kernel Fisher Discriminant Analysis (KFDA) chosen as fusion; finally, Nearest Neighbor (NN) classifier selected to perform recognition. Experimental results on ORL (Olivetti Research Laboratory) database CASIA (Chinese Academy of Sciences, Institute Automation) show that the dimension reduced, classified information utilized, correct recognition...

In widely linear beamforming, the second‐order non‐circularity property exhibited by signal‐of‐interest (SOI) can be exploited to improve interference and noise suppression. Assuming known SOI steering vector spatially white noise, several algebraic schemes for estimating coefficient are proposed in context of narrowband beamforming without training/reference signal. These new estimators based on signal purification, pseudoinverse, oblique projection, pre‐processed subspace separation....

A single period two echelon closed-loop supply chain system with retail competing was considered in this paper. Through considering the collecting price, wholesale price and prices as triangular fuzzy numbers, optimal decisions for manufacturer retailers explored by using theory game theory. The expressions of stackelberg equilibrium were established. By numerical examples, parameter analysis given.

Effective in silico compound prioritization is critical to identify promising candidates the early stages of drug discovery. Current methods typically focus on ranking based one single property, for example, activity, against a target. However, selectivity also key property that should be deliberated simultaneously so as reduce likelihood undesired side effects future drugs. In this paper, we present novel machine learning differential method dCPPP. This dCPPP learns models rank active...

Abstract In order to solve the problem that conventional tracker is not adapted abrupt motion, a tracking algorithm based on improved salp swarm (ISSA) was proposed. Visual considered be process of locating optimal position through interaction between leaders and followers in successive images. Firstly, adaptive mechanism leader follower introduced into original (SSA) balance exploitation exploration algorithm. This method can improve accuracy effect tracking. Secondly, golden‐sine used...

Rank criteria, a conventional recognition method of error correcting codes, was widely used for noiseless channel before. In this paper, the binary linear block code is considered, and it different from previous works. Firstly, an improved algorithm based on rank calculating to recognize length proposed. Then, probability statistics synchronization deviation used. At last, parity check matrix recovered by performing elementary row transformation dual vectors. The analysis confirmed...

Lion swarm optimization (LSO) inspired by the natural division of labor among lion king, lionesses and cubs in a group, i.e., king guarding, hunting following, is relatively novel intelligent technique. Due to its remarkable performance, canonical LSO has been extensively researched. However, how balance contradictions between exploration exploitation alleviate premature convergence are two critical concerns that need be dealt with study. To address these drawbacks, enhance broaden...