A major assumption in data mining and machine learning is that the training set test come from same domain. They share feature space have distribution. However, many real-world applications, usually different domains. Thus, there might be negative similarities between domains so transfer problem caused by similarity may happen. In this paper, we propose a novel method named active (ATL) to solve above problem. Specifically, orthogonal projection matrix weight coefficient vector are...

Age estimation from face images has attracted much attention due to its favorable and many real-world applications such as video surveillance social networking. However, most existing studies usually learn a single kind of age feature ignore other appearance features gender race, which have great influence on the pattern. In this paper, we proposed compact multifeature learning fusion method for estimation. Specifically, first used three subnetworks gender, information. Then, fused these...

As social media faces with large amounts of data and multimodal properties, cross-modal hashing (CMH) retrieval gains extensive applications its high efficiency low storage consumption. However, there are two issues that hinder the performance existing semantics-learning-based CMH methods: 1) exist some nonlinear relationships, noises, outliers in data, which may degrade learning effectiveness a model; 2) complementary relationships between label semantics sample be inadequately explored. To...

For multi-view face alignment, we have to deal with two major problems: 1) the problem of multi-modality caused by diverse shape variation when view changes dramatically; 2) varying number feature points self-occlusion. Previous works used nonlinear models or based methods for alignment. However, they either assume all are visible apply a set discrete separately without uniform criterion. In this paper, propose unified framework solve in which, both and variable modeled Bayesian mixture...

Domain adaptation (DA) aims to find a subspace, where the discrepancies between source and target domains are reduced. Based on this classifier trained by labeled samples can classify unlabeled well. Existing approaches leverage Graph Embedding Learning explore such subspace. Unfortunately, due 1) interaction of consistency specificity samples, 2) joint impact degenerated features incorrect labels in existing might assign unsuitable similarity, which restricts their performance. In paper, we...

Hashing has attracted widespread attention in the field of supervised cross-modal retrieval due to its advantages search and storage. However, there are still some issues be addressed, e.g.: 1) how effectively combine sample label semantics learn hash codes; 2) reduce high computational requirements brought by computing a pairwise similarity matrix; 3) solve discrete optimization problems. To cope with them, fast asymmetric hashing (FADCH) method is proposed this article. First, matrix...

Most existing domain adaptation (DA) methods aim to explore favorable performance under complicated environments by sampling. However, there are three unsolved problems that limit their efficiencies: i) they adopt global sampling but neglect exploit and local simultaneously; ii) either transfer knowledge from a perspective or perspective, while overlooking transmission of confident both perspectives; iii) apply repeated during iteration, which takes lot time. To address these problems,...

Selection of suppliers is the precondition and foundation supply chain operation. It an important aspect to choose best supplier for management. During recent years, how determine suitable in has become a key strategic consideration. However, nature these decisions usually complex unstructured. The proposed methodology consists two parts: 1) RST fairly new developed dealing with imprecise, uncertain, vague information. We assure weight selection based on Rough Set Theory (RST) model. 2)...

Direction information has been intensively investigated for Finger-Knuckle-Print (FKP) recognition. However, most existing direction-based KFP recognition methods are handcrafted, which heuristic and require too much prior knowledge to engineer them. In this paper, we propose a discriminative direction binary feature learning (DDBFL) method FKP We first convolution difference vector (DCDV) better describe the of images. Then, learn projection convert DCDV into codes, compact intra-class...

Automatic Modulation Classification (AMC) of communication signals plays a significant role in systems. However, conventional methods modulation classification have poor performance shallow water environment. Recently, the Stockwell-transform (S-transform), new time-frequency analysis method, receives widely attention different areas. In this paper, we introduce S-transform into and propose novel method under underwater acoustic channel. Firstly, set up model channel based on Bellhop...

One of the main difficulties in machine learning is how to solve large-scale problems effectively, and labeled data are limited fairly expensive obtain. In this paper a new semi-supervised SVM algorithm proposed. It applies tri-training improve SVM. The makes use large number unlabeled modify classifiers iteratively. Although doesn't put any constraints on classifier, proposed method uses three different SVMs as classification algorithm. Experiments UCI datasets show that can accuracy...

We propose a method to video segmentation via active learning. Shot is an essential first step segmentation. The color histogram-based shot boundary detection algorithm one of the most reliable variants algorithms. It not unreasonable assume that content does change rapidly within but across shots. Thus, we present metric based on blocked histogram (BCH) for inter-frame difference. Our normalized intersection BCH between contiguous frames. Hard cuts and gradual transitions can be detected as...

Corticosteroid has been proved to be one of the few effective treatments for COVID-19 patients. However, not all patients were suitable corticosteroid therapy. In this study, we aimed propose a machine learning model forecast response therapy in We retrospectively collected clinical data about 666 receiving between January 27, 2020, and March 30, from two hospitals China. The was evaluated by hospitalization time, oxygen supply duration, outcomes Least Absolute Shrinkage Selection Operator...

In recent years, multi-label learning has received a lot of attention. However, most the existing methods only consider global label correlation or local correlation. fact, on one hand, both and l... | Find, read cite all research you need Tech Science Press

Multi-agent systems are historically one of the most crucial problems and now become an important topic in cognitive theory. In view ontology theory, multi-agent usually related to roles agents executors roles. We investigate improved system by solving RBAC role-permission assignment problem with conflicting constrains this paper, under E-CARGO model that provides formalization, description conflicts algorithms used paper. Simulation experiments comprehensive consideration can verify...