[신간의 별자리x] 우리/미술, 그리고 ‘슬픔의 박물관’

Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

Sidechain technologies in blockchain networks: An examination and state-of-the-art review

Blockchain is an immutable type of distributed ledger that is capable of storing data without relying on a third party. Blockchain technology has attracted significant interest in research areas, including its application in the smart grid for cybersecurity. Although significant efforts have been devoted to utilizing blockchain in the smart grid for cybersecurity, there is a lack of comprehensive survey on blockchain in the smart grid for cybersecurity in both application and technological perspectives. To fill this gap, we conducted a comprehensive survey on blockchain for smart gird cybersecurity. This conducted survey presents the latest insights of ideas, architectures, and techniques of implementation that are relevant to blockchain's application in the smart grid for cybersecurity. This article aims at providing helpful guidance and reference for future research efforts specific to blockchain for cybersecurity in the smart grid.

Blockchain for Cybersecurity in Smart Grid: A Comprehensive Survey

In this paper, the performance of the proposed Convolutional Neural Network (CNN) with three well-known image recognition methods such as Principal Component Analysis (PCA), Local Binary Patterns Histograms (LBPH) and K–Nearest Neighbour (KNN) is tested. In our experiments, the overall recognition accuracy of the PCA, LBPH, KNN and proposed CNN is demonstrated. All the experiments were implemented on the ORL database and the obtained experimental results were shown and evaluated. This face database consists of 400 different subjects (40 classes/ 10 images for each class). The experimental result shows that the LBPH provide better results than PCA and KNN. These experimental results on the ORL database demonstrated the effectiveness of the proposed method for face recognition. For proposed CNN we have obtained a best recognition accuracy of 98.3 %. The proposed method based on CNN outperforms the state of the art methods.

A new method for face recognition using convolutional neural network

Nowadays, the development of the Internet of Things is growing rapidly. This condition raises security problems because of the many violations of security policies. In addition, the development of blockchain is also growing rapidly since it was first popularized in Bitcoin. Security issues on this IoT can be solved using blockchain. One way that can be done is to make secure communication between IoT devices. In this study, an IoT system will be created without and using blockchain which will then be compared between the two. The communication protocol used in IoT systems without blockchain is MQTT. Meanwhile, the blockchain platform used is Ethereum along with a smart contract. Both of these IoT systems will be analyzed for their security level by simulating attacks and observing their security aspects. The test results show that the IoT system based on blockchain technology has a higher level of security than the IoT system without blockchain technology.

Secure IoT Communication using Blockchain Technology

Human depicts their emotions through facial expression or their way of speech. In order to make this process possible for a machine, a training mechanism is needed to give machine the ability to recognize human expression. This paper compare and analyse the performance of three machine learning algorithm to do the task of classifying human facial expression. The total of 23 variables calculated from the distance of facial features are used as the input for the classification process, with the output of seven categories, such as: angry, disgust, fear, happy, neutral, sad, and surprise. Some test cases were made to test the system, in which each test cases has different amount of data, ranging from 165–520 training data. The result for each algorithm is quite satisfying with the accuracy of 75.15% for K-Nearest Neighbor (KNN), 80% for Support Vector Machine (SVM), and 76.97% for Random Forests algorithm, tested using test case with the smallest amount of data. As for the result using the largest amount of data, the accuracy is 98.85% for KNN, 90% for SVM, and 98.85% for Random Forests algorithm. The training data for each test case was also classified using Discriminant Analysis with the result 97.7% accuracy.

Comparative analysis of machine learning KNN, SVM, and random forests algorithm for facial expression classification

This study presents the real time implementation of object detection and tracking algorithm on mobile soccer robot. Object detection is considered as one of the most important task because ball and goal post are the main component in soccer. The system uses the combination of color-based segmentation and feature detection to detect the color and also the shape feature of the object used in the soccer robot game. The color segmentation uses thresholding method in Hue, Saturation, and Value (HSV) color space to differentiate the ball and goal post color from other objects in the field. Then, morphological operation is applied to the thersholded image to minimize the error. After that, Hough line transform is applied to detect the feature of the goal post. Then, ellipse detection is also applied to find the ball feature. This step is used so the desired object is correctly detected, not other object that have the same color. The final step is to calculate the image moments to determine the centroid of the objects and tracking it. Object's color, feature, and coordinate are obtained from this purposed method. In the implementation, the robot has successfully detect ball, goal post, and its position in a real time manner.

Color-based segmentation and feature detection for ball and goal post on mobile soccer robot game field

Blockchain is a distributed database that has advantage of being irreversible, transparent and decentralized. However, beside of its advantages, blockchain still face several challenges, one of them is scalability. One of the scalability challenges is increasing block size for every new transaction. The size of blockchain reached more than 180 GB. The block size will burden the node in conducting transaction verification. To overcome this problem, this paper has an objective to make summary blocks and compression for the summary blocks so that it can increase the space saving of blockchain and make it easier for nodes to verify transactions. The proposed method used for transactions whose entities can be transferred. This method combines block summarization algorithm and deflate compression algorithm. The proposed method applied to Bitcoin blockchain. Space saving is calculated by comparing the total size of original block with the total size of block that has been summarized and the total size of summary block which has been compressed. From the experiment results, the value of space saving for summary blocks is 22.318% and space saving for summary blocks that have been compressed is 78.104%.

Block Summarization and Compression in Bitcoin Blockchain

Blockchain technology has gained immense popularity because many researchers believe that it could solve numerous problems and could be applied in various fields of study. Unfortunately, behind its potentials, blockchain also possessed many challenges and limitations. The highlighted problem is the usability aspect of blockchain technology examined from developer and user perspective. This paper tried to address this problem by proposing query functionalities, with the help of query layer system, to facilitate the developer and the user to access blockchain data easily. There are three main query functionalities that will be discussed in this paper: (1) finding blockchain data based on multiple search parameters (retrieval query), (2) providing simple statistical analysis from a collection of blockchain data (aggregate query) and (3) sorting blockchain data according to its blockchain component (ranking query). For the implementation stage, Ethereum is used as platform to provide blockchain network, MongoDB is used as cloud storage service and REST API is used as web services. For the testing stage, throughput and time response are used to evaluate the performance of the developed query functionalities in the query layer system. The results are: (1) the throughput of query layer system is lower than Ethereum service for blockchain data retrieval and (2) the time response of query layer system is affected by the number of thread and the amount of data stored in cloud storage.

Kusprasapta Mutijarsa

Papers

Secure IoT Communication using Blockchain Technology

Comparative analysis of machine learning KNN, SVM, and random forests algorithm for facial expression classification

Color-based segmentation and feature detection for ball and goal post on mobile soccer robot game field

Block Summarization and Compression in Bitcoin Blockchain

Query Support for Data Processing and Analysis on Ethereum Blockchain