Abstract Transfer learning has benefited many real‐world applications where labeled data are abundant in source domains but scarce the target domain. As there usually multiple relevant knowledge can be transferred, transfer ( MSTL ) recently attracted much attention. However, we facing two major challenges when applying . First, without about difference between and domains, negative occurs is transferred from highly irrelevant sources. Second, existence of imbalanced distributions classes,...

Transfer learning has benefitted many real-world applications where labeled data are abundant in source domains but scarce on the target domain. As there usually multiple relevant knowledge can be transferred, Multiple Source Learning (MSTL) recently attracted much attention. Most existing MSTL methods work an offline fashion that they have to store all domain before learning. However, some time-critical arrive sequentially large volume, a fast and scalable online method transfer from is...

In recent years, information trustworthiness has become a serious issue when user-generated contents prevail in our world. this paper, we investigate the important problem of estimating from perspective correlating and comparing multiple data sources. To certain extent, consistency degree is an indicator reliability--Information unanimously agreed by all sources more likely to be reliable. Based on principle, develop effective computational approach identify consistent Particularly, analyze...

As is known to all, traditional clustering algorithms do not work well due the topological features of protein-protein interaction networks. An improved method based on bacteria foraging optimization (BFO) mechanism and intuitionistic fuzzy set, short for BFO, proposed in this paper, which trigonometric function used define membership degrees indeterminacy degree introduced detect overlapping modules. In chemotactic operation algorithm initializes a cluster center according comprehensive...

In this paper, spectral approximations for distributed optimal control problems governed by the Stokes equation are considered.And constraint set on velocity is stated with L 2 -norm.Optimality conditions of continuous and discretized systems deduced Karush-Kuhn-Tucker a Lagrange multiplier depending constraint.To solve equivalent high accuracy, Galerkin employed to discretize constrained systems.Meanwhile, we adopt parameter λ in pressure approximation space, which also guarantees inf-sup...

Abstract The distributed model predictive control (MPC) is studied for the tracking and formation problem of multi‐agent system with time‐varying communication topology. At each sampling instant, agent solves an optimization respecting input state constraints, to obtain its optimal input. In cost function agent, weighting coefficient properly updated so that adverse effect topology on closed‐loop stability can be counteracted. It shown overall achieve desired objectives. effectiveness...

This paper presents a stochastic approach to describe input uncertainties and their propagation through the nonlinear shallow-water equations. The formulation builds on finite-volume model with Godunov-type scheme for its shock capturing capabilities. Orthogonal polynomials from Askey provide expansion of variables in terms finite number modes which mean higher-order moments distribution can be derived. orthogonal property allows use Galerkin projection derive separate equations individual...

Previous chapter Next Full AccessProceedings Proceedings of the 2013 SIAM International Conference on Data Mining (SDM)On Handling Negative Transfer and Imbalanced Distributions in Multiple Source LearningLiang Ge, Jing Gao, Hung Ngo, Kang Li, Aidong ZhangLiang Zhangpp.261 - 269Chapter DOI:https://doi.org/10.1137/1.9781611972832.29PDFBibTexSections ToolsAdd to favoritesExport CitationTrack CitationsEmail SectionsAboutAbstract learning has benefited many real-world applications where labeled...

We investigate how to estimate information trustworthiness by considering multiple sources jointly in a latent matrix space. particularly focus on user review and recommendation systems, as there are platforms where people can rate items services that they have purchased, many potential customers rely these opinions make decisions. Information is serious problem because ratings generated freely end-users so stammers take advantage of freedom speech promote their business or damage reputation...

Link prediction is an important task in social networks and data mining for understanding the mechanisms by which form evolve. In most link researches, it assumed either a snapshot of network or with some missing links available. Most existing researches therefore approach this problem exploring topological structure using only one source information. However, many application domains, addition to interest, there are number auxiliary information work, we introduce pseudo cold start multiple...

Knowing students' personality traits helps to improve the environment of learning and living in college. However, questionnaire commonly used for prediction has some inherent limitations such as inefficiency proneness cheating. In this paper, we propose a novel approach based on data mining determine particular extraversion introversion. Real set comes from 79 undergraduate students at Chongqing University. We use results Revised Eysenck Personality Questionnaire Short Scale Chinese...

This paper presents a stochastic approach to model input uncertainty with general statistical distribution and its propagation through the nonlinear long-wave equations. A Godunov-type scheme mimics breaking waves as bores for accurate description of energy dissipation in runup process. The polynomial chaos method expands flow parameters into series orthogonal modes, which contain properties space. spectral projection technique determines modes from ensemble averages systematically sampled...

Detecting functional modules from protein-protein interaction (PPI) networks is an active research area with many practical applications. However, there always a critical concern on the false PPI interactions which are derived high-throughput experiments and unsatisfactory results obtained single network severe information insufficiency. To address this problem, we propose Collective Non-negative Matrix Factorization (CoNMF) based soft clustering method efficiently integrates of gene...

Federated Learning (FL) enables plenty of edge computing devices to jointly learn a model without data sharing. However, in real-world federated datasets, traditional FL algorithms perform poorly with non-IID data. To better from data, we propose DMFL (Dynamic Margin for Learning) as strategy improve the performance learning highly skewed imbalanced dataset. Accordingly, theoretical analysis dynamic margin loss FL, demonstrating that it is applicable yield lower generalization error bound....

Abstract In this paper, we study the mathematical formulation for an optimal control problem governed by a linear parabolic integro-differential equation and present optimality conditions. We then set up its weak finite element approximation scheme. Based on these derive priori error estimates both in H 1 L 2 norms. Furthermore some numerical tests are presented to verify theoretical results.

With memory information as characteristic, a classified method to detect APT (Advanced Persistent Threat)Trojans in cloud computing is proposed this paper. Memory analysis and fuzzy C-means (FCM) algorithm based on the optimized initial cluster centers detects similarity of Trojans. Without influence normal operation virtual machine cloud, classifier can determine whether it malware or not. The overcome shortage feature scanning technology which could not recognize unknown Trojans,...

Password recovery of WPA2-PSK is an important problem in digital forensics. Since the encryption mechanism WPA-PSK gradually enhanced, it difficult to deal with this by traditional methods such as brute force, rainbow table, Markov model, and so on. In paper, we give a new method based on simulated annealing (SA) hidden markov model (HMM). The main principle create known password SA which could be used generate candidates wireless network recovery. It means that passwords are given...