T R A C K       P A P E R
Close

Login Panel

Close tab

Password Meter

Volume 07 Issue 09 (September 2020)

S.No. Title & Authors Page No View
1

Title : A Review of Consensus Algorithms in Blockchain

Authors : Lusheng Ji

Click Here For Abstract

Download Certificate
Abstract :

Consensus algorithm is the key to achieve the final consistency of blockchain system. The consensus algorithm of blockchain also provides a solution to the consensus problem in distributed system.After explaining the development history of consensus algorithm, this paper analyzes the current mainstream block chain consensus algorithm model as well as its advantages and disadvantages.And the development of block chain consensus algorithm in the future is prospected

1-2
2

Title : Image Aesthetic Assessment Based on Deep Learning: An Survey

Authors : Yaoting Wang

Click Here For Abstract

Download Certificate
Abstract :

With the rise of deep learning in recent years, the field of image aesthetics quality evaluation has developed from traditional machine learning methods to end-to-end convolutional neural network evaluation methods, which has achieved a qualitative leap in the results of image aesthetics quality evaluation. This article mainly summarizes and introduces the research of convolutional neural network methods in image aesthetics evaluation in recent years. It aims to solve the problems of incomplete generalization and insufficient understanding of the existing review literature. Explains in detail the development from manual feature extraction to deep learning, image aesthetics related data sets, and various application directions of image aesthetics evaluation, including automatic image cropping based on image aesthetics, image semantic line detection, image composition classification, and image aesthetic attributes Analysis etc. Finally, the future work in the direction of image aesthetics is prospected

3-6
3

Title : Research on Discovering Time Series Motifs Based on Stacked Auto Encoder

Authors : YiHong Gao, XinMing Duan, ZiLiang Chen

Click Here For Abstract

Download Certificate
Abstract :

Time series are widely used in financial data, streaming media data, weather data, census data, and system log data. It is very important to find frequently repeated patterns (motifs) in the time series. But, finding motifs is a complicated task due to its huge dimensions. In order to fix the dimension problems and reduce the calculation time of time series, researchers have done a lot of research on the dimensions of the time series, but they have not made much breakthrough. Therefore, this paper has carried out related work research to improve the problem: (i) Preprocessing the time series.(ii) Using the more popular neural network-stacked autoencoder to extract features of time series, which can reduce the number of time series calculations. (iii) Running a large number of experiments to verified the accuracy of time series motif search combined with stacked autoencoders.

The study found that the method in this paper can not only guarantee the validity of the time series motifs, but also ensure the accuracy (about 88%).

7-11
4

Title : A Review of Aerial Images Object Detection Based on Deep Learning

Authors : Xuechun Wang

Click Here For Abstract

Download Certificate
Abstract :

Due to the close relationship between object detection and image understanding, it has attracted a lot of research attention in recent years. Driven by deep learning, the problem of object detection has been developed rapidly, especially in natural scenes, there have been a series of breakthroughs, but the progress in remote sensing images has been slow. Due to the fact that the detection object of the aerial image is generally small, the object may be rotated in the picture, and the detection instance is large in magnitude. As a result, the existing object detection algorithm directly used in the aerial images object detection effect is not ideal. This article first introduces several popular object detection algorithms and analyzes the characteristics of each algorithm. Secondly, the characteristics of the aerial images data set are introduced, and the existing aerial images object detection algorithms are analyzed and summarized. Finally, discuss the existing problems and some insights on future object detection work

12-14
5

Title : Human Resource Information System

Authors : Nisha R

Click Here For Abstract

Download Certificate
Abstract :

There has been a tremendous change in the field of human resource management is playing a very dominant role in organizations with newly evolved strategies and systems that keeps evolving with the introduction of technology. Human resource functions are mostly connected with the employees, stake holders and the people who are connected with the firm. It is mainly designed in such a manner that the individual goals align with the organizational goals and which in turn reflects the performance of the work force and productivity of the concern.

15-18
6

Title : Passivity Analysis of Time-Delayed Neural Networks with both Leakage Delay and Randomly Occurring Uncertainties

Authors : Yanyu Wang

Click Here For Abstract

Download Certificate
Abstract :

This article studies the related issues of the robustness and passiveness of neural networks with time-varying delays and leakage delays with parameter uncertainties.The white noise sequence that obeys the relevant Bernoulli distribution enters the system in randomly form.By choosing the appropriate LKFs, and using methods such as Wirtinger inequality and free weight matrix to improve the delay standard, and express it in the form of linear matrix inequality.Sufficient conditions are established to ensure the robust random stability and passivity of the neural network under consideration.Finally, a simulation example is given using the LMI toolbox to prove the validity and conservativeness of the standard proposed in this article

19-23
7

Title : A Survey of Image Inpainting Research Based on Generative Adversarial Network

Authors : Youyu Sun, Baoshan Sun

Click Here For Abstract

Download Certificate
Abstract :

Purpose The concept of image restoration originated from the manual restoration of murals and other artworks during the European Renaissance. Image restoration technology is different from other processing technologies in the field of computer vision. It has high requirements for image feature extraction. Therefore, restoration techniques based solely on convolution and other methods cannot achieve human visual recognition in terms of restoration effects. The proposal of Generative Adversarial Network (GAN) provides a new idea for the field of image restoration. It adopts the idea of image generation model and discriminant model for adversarial training, and the repair effect is more in line with the characteristics of visual perception. Method Generative confrontation network has more powerful feature learning and feature expression capabilities than traditional machine learning algorithms when performing image processing. In the early stage of a large number of literature research work, it is found that the conditional generative confrontation network CGAN, the deep convolution-based generative confrontation network DCGAN and the Wasserstein generative confrontation network are more widely used. This article mainly introduces the basic ideas and methods of GAN, CGAN, DCGAN and WGAN, and analyzes and summarizes their advantages and disadvantages in image restoration. Conclusion The current research on GAN-based image restoration methods has made a certain degree of progress, but GAN as a new type of network model still needs further research in theory

24-27