Distant Supervised Relation Extraction with Cost-Sensitive Loss

Recently, many researchers have concentrated on distant supervision relation extraction (DSRE). DSRE has solved the problem of the lack of data for supervised learning, however, the data automatically labeled by DSRE has a serious problem, which is class imbalance. The data from the majority class obviously dominates the dataset, in this case, most neural network classifiers will have a strong bias towards the majority class, so they cannot correctly classify the minority class. Studies have shown that the degree of separability between classes greatly determines the performance of imbalanced data. Therefore, in this paper we propose a novel model, which combines class-to-class separability and cost-sensitive learning to adjust the maximum reachable cost of misclassification, thus improving the performance of imbalanced data sets under distant supervision. Experiments have shown that our method is more effective for DSRE than baseline methods.     

Face expression recognition system based on ripplet transform type II and least square SVM

Multimedia Tools and Applications - This paper discusses the development of an efficient and automated system for the recognition of facial expressions, which is essentially an application...     

Metabolic Fate of the Environmental Mutagen 3-Nitrofluoranthene in the Rat

3-Nitrofluoranthene, a potent environmental mutagen, was administered to rats intraperitoneally at 10 mg/kg, labelled with ¹⁴C. Urine contained 15 to 20% of the dose, eliminated mainly within the first 24 hours after dosing. Over 95% of the radioactivity in unhydrolysed urine chromatographed with the solvent front in a reverse-phase HPLC system, indicating extensive conjugation. After hydrolysis with β-glucuronidase (containing sulfatase) and concentration on C18-Sep-Pak, urine was fractionated by HPLC for further characterization. NMR analysis of the metabolite fractions indicated that positions 4, 8 and 9 were major sites of oxidation. Both acetamide and amine derivatives were formed. The finding of multiple oxidation and incomplete acetylation is in contrast to the fate of 1-nitropyrene, where 1-acetamidopyren-6-ol, itself a potent mutagen, represents the major metabolic product. This difference in metabolism pathways may have implications for the genotoxicity of 3-nitrofluoranthene.     

Mechanisms of Solution Reactions of Cyclo-Penta[cd]Pyrene Oxide and Acenaphthylene Oxide

Rates and products of the acid-catalyzed and spontaneous reactions of cyclopenta[cd]pyrene oxide (2) and acenaphthalene oxide (3) in 1:3 dioxane-water and water, respectively, have been determined. Both 2 and 3 undergo acid-catalyzed hydrolyses to yield cis diol as the major product, and undergo spontaneous reactions to yield predominantly ketone products. The mechanisms of these reactions are discussed.     

A Deep Learning Approach to Detection and Mitigation of Distributed Denial of Service Attacks in High Availability Intelligent Transport Systems

Mobile Networks and Applications - In the era of Internet of Things (IoT) powered by 5G technologies, Automobile Industry is headed towards a revolution. In Intelligent Transport Systems (ITS),...     

Single image super resolution for texture images through neighbor embedding

This article proposes an improved learning based super resolution scheme using manifold learning for texture images. Pseudo Zernike moment (PZM) has been employed to extract features from the texture images. In order to efficiently retrieve similar patches from the training patches, feature similarity index matrix (FSIM) has been used. Subsequently, for reconstruction of the high resolution (HR) patch, a collaborative optimal weight is generated from the least square (LS) and non-negative matrix factorization (NMF) methods. The proposed method is tested on some color texture, gray texture, and some standard images. Results of the proposed method on texture images advocate its superior performance over established state-of-the-art methods.

     

A Real-Time and Ubiquitous Network Attack Detection Based on Deep Belief Network and Support Vector Machine

In recent years, network traffic data have become larger and more complex, leading to higher possibilities of network intrusion. Traditional intrusion detection methods face difficulty in processing high-speed network data and cannot detect currently unknown attacks. Therefore, this paper proposes a network attack detection method combining a flow calculation and deep learning. The method consists of two parts: a real-time detection algorithm based on flow calculations and frequent patterns and a classification algorithm based on the deep belief network and support vector machine (DBN-SVM). Sliding window (SW) stream data processing enables real-time detection, and the DBN-SVM algorithm can improve classification accuracy. Finally, to verify the proposed method, a system is implemented. Based on the CICIDS2017 open source data set, a series of comparative experiments are conducted. The method’s real-time detection efficiency is higher than that of traditional machine learning algorithms. The attack classification accuracy is 0.7 percentage points higher than that of a DBN, which is 2 percentage points higher than that of the integrated algorithm boosting and bagging methods. Hence, it is suitable for the real-time detection of high-speed network intrusions.      

A password based authentication scheme for wireless multimedia systems

In this digital era, where Internet of Things (IoT) is increasing day by day, use of resource constrained devices is also increasing. Indeed, the features such as low cost, less maintenance, more adaptive to hostile environment, etc. make the wireless multimedia devices to be the best choice as the resource constrained devices. For the security, the end user device requires to establish the session key with the server before transferring the data. Mobile is one of the device having more and more usage as wireless multimedia device in recent years. In 2013, Li et al. proposed an efficient scheme for the wireless mobile communications and claimed it to be secure against various attacks. Recently, Shen et al. claimed that the scheme of Li et al. is still vulnerable to the privileged insider attack, the stolen verifier attack and finally proposed a scheme to withstand the mentioned and other attacks. However, in this paper we claim that the scheme of Shen et al. is still susceptible to the user anonymity, the session specific temporary information attack and the replay attack. In addition, Shen et al.’s scheme requires more time due to many operations. Further, we propose an efficient scheme that is secure against various known attacks and due to reduced time complexity our scheme is a preferred choice for the wireless mobile networks and hence for wireless multimedia systems.     

Evidence-based personal applications of medical computing models in risk factors of cardiovascular disease for the middle-aged and elderly

Medical applications on cardiovascular disease (CVD) for hybrid computing models are an emerging research area. The CVD, including stroke, hypertension, and high cholesterol, is one of 10 leading causes of death in Taiwan in middle-aged and elderly; in particular, the CVD has become the top killer in advanced countries. Thus, this serious but interesting issue triggers the study to focus on patients of the CVD. The study explores variables, influencing cardiovascular functions for four risk factors of blood pressure, blood glucose, blood fat, and kidney diseases, in the middle-aged and elderly. By the data collection of regular physical examination system from a regional hospital, the original dataset contains 52 variables collected from October 2011 to February 2014. We model a hybrid knowledge-based classification system to organize expert experiences, integrated linear and nonlinear attribute selection methods, data discretization of smart expert method, rough set theory, the LEM2 algorithm, and rule-filtering technique to classify the CVD for the early warning purpose. After data cleaning, 20 attributes with 2027 records are remained. For effectively identifying the variables of CVD subjects, this study reclassifies the above four risk diseases into three classes: no disease, 1&2 diseases, and 3&4 diseases. To verify performance of the proposed procedure, we experience an empirical experiment to compare the full 20 used attributes, the used attributes of integrated linear and nonlinear attribute selections with rule-filtering technique, and various classifiers. Conclusively, the 13 used attributes obtained from optimal accuracy become the key determinants that affect the four risk factors of the CVD. The empirical results and findings benefit doctors’ and medical institutions’ early medical recommendations and treatments with the advantages of significantly reducing morbidity of CVD.     

Deep Learning for Financial Engineering

Computational Economics -