Ebook: Machine Learning and Artificial Intelligence
Machine learning and artificial intelligence are already widely applied to facilitate our daily lives, as well as scientific research, but with the world currently facing a global COVID-19 pandemic, their capacity to provide an important tool to support those searching for a way to combat the novel corona virus has never been more important.
This book presents the proceedings of the International Conference on Machine Learning and Intelligent Systems (MLIS 2020), which was due to be held in Seoul, Korea, from 25-28 October 2020, but which was delivered as an online conference on the same dates due to COVID-19 restrictions. MLIS 2020 was the latest in a series of annual conferences that aim to provide a platform for exchanging knowledge about the most recent scientific and technological advances in the field of machine learning and intelligent systems. The annual conference also strengthens links within the scientific community in related research areas.
The book contains 53 papers, selected from more than 160 submissions and presented at MLIS 2020. Selection was based on the results of review and scored on: originality, scientific/practical significance, compelling logical reasoning and language. Topics covered include: data mining, image processing, neural networks, human health, natural language processing, video processing, computational intelligence, expert systems, human-computer interaction, deep learning, and robotics.
Offering a current overview of research and developments in machine learning and artificial intelligence, the book will be of interest to all those working in the field.
The COVID-19 pandemic which began in late 2019 has posed a historical challenge to society. At the time of writing, scientists, clinicians, and healthcare experts around the globe continue to search for new technologies to support those tackling the pandemic. The evidence deriving from the application of Machine Learning (ML) and Artificial Intelligence (AI) to previous epidemics encourages researchers by providing a new angle with which to fight the novel corona virus outbreak. ML and AI are widely applied to facilitate our daily lives, as well as in scientific research.
The International Conference on Machine Learning and Intelligent Systems (MLIS) is an annual conference that aims to provide a platform for exchanging knowledge about the most recent scientific and technological advances in the field of ML and intelligent systems. It also strengthens links within the scientific community in related research areas. MLIS 2020 was initially scheduled to be held in Seoul, Korea from 25–28 October 2020, but was changed to an online conference on the same dates due to COVID-19 restrictions. Scientific topics covered at MLIS 2020 included data mining, image processing, neural networks, human health, natural language processing, video processing, computational intelligence, expert systems, human-computer interaction, deep learning, robotics, etc.
This book contains 53 papers selected from more than 160 submissions and presented at MLIS 2020. Selection was based on the results of review scored on: originality, scientific/practical significance, compelling logical reasoning and language. Rigorous peer review was conducted by members of the Editorial Committee led by Guest Editors Prof. Antonio J. Tallón-Ballesteros and Prof. Chi-Hua Chen, as well as active reviewers like Prof. Dheyaa Jasim Kadhim, Prof. Saeid Seyedi, Prof. Ivan Izonin, Prof. Behrouz Pourghebleh, Prof. Tahereh Eftekhari, Prof. Ilaria Cacciari, Prof. Ogundokun Roseline Oluwaseun, who engaged authors to answer stimulating questions for the camera-ready versions of their papers. We believe the 53 papers selected will provide some insights in the field of ML and AI.
On behalf of the conference organizers, we would like to take this opportunity to express our sincere thanks to the Guest Editor and all the Reviewers for their tremendous efforts and dedication to the conference. We also wish to thank all the authors for their relevant contributions to the conference, as well as all our colleagues at the publishers IOS Press for their support and tireless efforts in preparation for the publication of the conference proceedings. We believe that with such earnest support and contributions, future MLIS Conferences will also scale new heights.
Antonio J. Tallón-Ballesteros
University of Huelva (Spain)
Seville city, Spain
Fuzhou University (China)
Fuzhou City, Fujian, China
The problem of influence maximization (IM) represents a major challenge for modern network science, with direct applicability in political science, economy, epidemiology, and rumor spreading. Here, we develop a novel computational intelligence framework (GenOSOS) based on genetic algorithms with emphasis on the optimal layout of spreader nodes in a network. Our algorithm starts with solutions consisting of randomly selected spreader nodes; then, by defining custom original crossover and mutation operators, we are able to obtain, in a short number of genetic iterations, nearly optimal solutions in terms of the nodes’ topological layout. Experiments on both synthetic and real-world networks show that the proposed GenOSOS algorithm is not only a viable alternative to the existing node centrality approach, but that it outperforms state of the art solutions in terms of spreading coverage. Specifically, we benchmark GenOSOS against graph centralities such as node degree, betweenness, PageRank and k-shell using the SIR epidemic model, and find that our solution is, on average, 11.45% more efficient in terms of diffusion coverage.
Major depressive disorder is one of common mental disorders globally. It is best to be early detected and cured. This work introduces a method to detect depressive disorder at risk via a behavior made on Facebook platform. A set of features related to Facebook main functions including amount of posting, sharing, commenting and replying is designed to represent users’ activities in a numerical value form. The collected data with periodic and consecutive aspects are gathered without interpreting content. Thus, the data are easier to be collected with less privacy issue. To distinct between positive and negative depression-at risk, PHQ-9 questionnaire, a standard tool commonly used to screen depression patient in Thailand, was used. These features hence are used in supervised learning classification algorithm for detecting a risk of being depressive disorder. From the experiment of 160 Thai Facebook users, the statistical result indicated that depression-positive users tend to do consecutive actions and rarely reply to other comments. Moreover, they often have activities in late night. The classification experiment shows that the designed features based on users’ activities from Facebook with deep learning algorithm yields about 87% in terms of F-measure. After analyzing the data, we thus split data regarding users’ gender and removed obviously low active data, and the F-measure from classification went up to 91.4 which improves for 4 points.
Classification is a data mining task and which is a two-phase process: learning and classification. The learning phase consists of constructing a classifier or a model from a labeled set of objects. The classification phase consists classifying new objects by using the generated classifier. Different approaches have been proposed for supervised classification problems through Formal Concept Analysis, and which is a mathematical theory to build upon hierarchies of formal concepts. The proposed approaches in literature rely on the use of single classifier and ensemble methods. Single classifier methods vary between them according to different criteria especially the number of formal concepts generated. We distinguish overall complete lattice methods, sub-lattice methods and concept cover methods. Methods based on ensemble classifiers rely on the use of many classifiers. Among these methods, there are methods based on sequential training and methods based on parallel training. However, with the large volume of data generated from various sources, the process of knowledge extraction with traditional methods becomes difficult. That’s why new methods based on distributed classifier have recently appeared. In this paper, we present a survey of many FCA-based approaches for classification by dividing them into methods based on a mono-classifier, methods based on ensemble classifiers and methods based on distributed classifiers. Different methods are presented and compared within this paper.
Edge-preserving and structure-preserving smoothing filtering has attracted much interest in the last decades. A conventional linear filter effectively smoothens noise in homogeneous regions but blurs the edges of an image. This study aimed to present an adaptive guided filter using a cross-based framework. The proposed method outperformed many other algorithms in terms of sharpness enhancement and noise reduction. Moreover, the cross-based adaptive guided filter had a fast and nonapproximate linear-time algorithm as the guided filter.
We propose a deep learning framework for anisotropic diffusion which is based on a complex algorithm for a single image. Our network can be applied not only to a single image but also to multiple images. Also by blurring the image, the noise in the image is reduced. But the important features of objects remain. To apply anisotropic diffusion to deep learning, we use total variation for our loss function. Also, total variation is used in image denoising pre-process. With this loss, our network makes successful anisotropic diffusion images. In these images, the whole parts are blurred, but edge and important features remain. The effectiveness of the anisotropic diffusion image is shown with the classification task.
This study aims to describe a model that will apply image processing and traditional machine learning techniques specifically Support Vector Machines, Naïve-Bayes, and k-Nearest Neighbors to identify whether or not a given breast histopathological image has Invasive Ductal Carcinoma (IDC). The dataset consisted of 54,811 breast cancer image patches of size 50px x 50px, consisting of 39,148 IDC negative and 15,663 IDC positive. Feature extraction was accomplished using Oriented FAST and Rotated BRIEF (ORB) descriptors. Feature scaling was performed using Min-Max Normalization while K-Means Clustering on the ORB descriptors was used to generate the visual codebook. Automatic hyperparameter tuning using Grid Search Cross Validation was implemented although it can also accept user supplied hyperparameter values for SVM, Naïve Bayes, and K-NN models should the user want to do experimentation. Aside from computing for accuracy, the AUPRC and MCC metrics were used to address the dataset imbalance. The results showed that SVM has the best overall performance, obtaining accuracy = 0.7490, AUPRC = 0.5536, and MCC = 0.2924.
The main purpose of disaggregation is to decompose a signal into a set of other signals that together constitute it. This approach could be applied to audio signals, health care, home automation, ubiquitous systems and energy systems. It may be unworkable to individually measure the energy consumption of loads in a system simultaneously and, through disaggregation, we can make an inference using a main meter. The main contribution of this work is to use PCA to extract representativeness of an energy consumption signal we want to disaggregate, identifying its most relevant characteristics. The field of study is relevant because it allows information to be obtained in a simpler and cheaper way about the individual consumption of loads that make up a system. This opens up perspectives for other approaches such as smart grids and IoT. We demonstrate that when compared to other techniques, the proposal produces more accurate disaggregation results.
For the five-axis machine into the singular region in the process of parts processing, resulting in a discontinuous and rapid rotation of the axis of rotation of large angles. Based on the analysis of the cause of the obvious ripple on the machined surface and the influence on the machining precision, a mathematical model of the singular region is established, and an optimization method of the tool path in the singular region is proposed. The simulation and practical machining results show that the method can effectively overcome the problem of excessive movement of the rotating shaft in the Song singular region of 5-axis machine tool, and solve the surface corrugated defects caused by the problem, while improving the processing efficiency.
Big Data in medicine contains conceivably fast processing of large data volumes, alike new and old in perseverance associate the diagnosis and treatment of patients’ diseases. Backing systems for that kind activities may include pre-programmed rules based on data obtained from the medical interview, and automatic analysis of test diagnostic results will lead to classification of observations to a specific disease entity. The current revolution using Big Data significantly expands the role of computer science in achieving these goals, which is why we propose a computer data processing system using artificial intelligence to analyse and process medical images. We conducted research that confirms the need to use GPUs in Big Data systems that process medical images. The use of this type of processor increases system performance.
Compared with ordinary large-scale structural parts, super-large aircraft aluminum alloy integral frame parts have the characteristics of large size, high ribs and thin-walled, which lead to the difficulty of deformation control and dimensional accuracy assurance in the machining process, and the problems of spring knife and broach are easy to occur. In this paper, the research on super-large aluminum alloy integral frame parts is carried out, and a set of methods with part deformation control and coordinate drift error control are proposed, and the processing programming strategy is further optimized. This method has been successfully applied to a super-large aircraft aluminum alloy integral frame part, which greatly reduces the deformation of parts, improves the processing stability, and improves the processing efficiency by about 30%.
We present an automated pipeline for the generation of synthetic datasets for six-dimension (6D) object pose estimation. Therefore, a completely automated generation process based on predefined settings is developed, which enables the user to create large datasets with a minimum of interaction and which is feasible for applications with a high object variance. The pipeline is based on the Unreal 4 (UE4) game engine and provides a high variation for domain randomization, such as object appearance, ambient lighting, camera-object transformation and distractor density. In addition to the object pose and bounding box, the metadata includes all randomization parameters, which enables further studies on randomization parameter tuning. The developed workflow is adaptable to other 3D objects and UE4 environments. An exemplary dataset is provided including five objects of the Yale-CMU-Berkeley (YCB) object set. The datasets consist of 6 million subsegments using 97 rendering locations in 12 different UE4 environments. Each dataset subsegment includes one RGB image, one depth image and one class segmentation image at pixel-level.
In this work, a fuzzy linear equation AX + B = 0, is solved, were A, B y C are triangular diffuse numbers, could also be trapezoidal. For this type of equations there are several solution methods, the classic method that does not always obtain solutions, the most used is the method of alpha cuts and arithmetic intervals that although it always finds a solution, as a value is taken closer to zero (more inaccurate), the solution satisfies less to the equation. The new method using the expected interval, allows to obtain a smaller support set where the solutions come closer to satisfying the equation, also allows to find a single interval where the best solutions for decision making are expected to be found. It is recommended to study the incorporation of the concept of the expected interval in the methods to solve systems of fuzzy linear equations
Currently, the software handled by hackers is the main one to tackle a series of empirical knowledge, with this software attacking and helping organizations. The main objective is to analyze and systematize the software that is detected by hackers and crackers, in order to prevent risks and study the tactical levels and strategies for a given process. The analytical method is used in this investigation, for the study or analysis of the offensive software structure in public organizations. The results obtained from this research were an attack launching algorithm, software prototype taken by hackers, massive obfuscation model, and quantitative encryption model. It was concluded that piracy tools are used for preventive prevention and systems aggression, that is, to be defensive or offensive for a period, throughout an attack cycle.
This research work proposes a Facial Emotion Recognition (FER) system using deep learning algorithm Gated Recurrent Units (GRUs) and Robotic Process Automation (RPA) for real time robotic applications. GRUs have been used in the proposed architecture to reduce training time and to capture temporal information. Most work reported in literature uses Convolution Neural Networks (CNN), Hybrid architecture of CNN with Long Short Term Memory (LSTM) and GRUs. In this work, GRUs are used for feature extraction from raw images and dense layers are used for classification. The performance of CNN, GRUs and LSTM are compared in the context of facial emotion recognition. The proposed FER system is implemented on Raspberry pi3 B+ and on Robotic Process Automation (RPA) using UiPath RPA tool for robot human interaction achieving 94.66% average accuracy in real time.
With the rapid development of the Internet and big data, the data resources in various industries and technical fields are constantly emerging and growing. How to monitor and identify the effective data information in the massive big data has become one of the key contents of the current scientific and technological information work. This paper designs and implements the advanced technology monitoring system based on multi-source heterogeneous data. It comprehensively uses information collection technology, database technology and big data mining technology to realize the accurate monitoring, acquisition and analysis of multi-source heterogeneous data. It reveals the coupling relationship of technologies, people and institutions in different fields and the future technology development trend, and finally visualizes in various states It provides a reference for the strategic decision-making of relevant government departments, and provides efficient and convenient research tools and methods for scientific research institutes and enterprises.
In this work, the AutonomousSystems4D package is presented, which allows the qualitative analysis of non-linear differential equation systems in four dimensions, as well as drawing the phase surfaces by immersing R4 in R3. The package is programmed in the computational tool Octave. As a case study applied to the new Lorenz 4D System, sensitivity was found in the initial conditions, Lyapunov exponents, Kaplan Yorke dimension, a stable and unstable critical point, limit cycle, Hopf bifurcation, and hyperattractors. The package could be adapted to perform qualitative analysis and visualize phase surfaces to autonomous systems, e.g. Sprott 4D, Rossler 4D, etc. The package can be applied to problems such as: design, analysis, implementation of electronic circuits; to message encryption.
Logistic regression is widely used in decision problems to classify inputs through training from the previously known training data. In this paper, we propose an approach to detecting similar versions of software by learning with logistic regression on binary opcode information. Because the binary opcode information has detailed information for executing software on an individual machine, the learning from the binary opcode information can provide effective information in detecting similar versions of software. To evaluate the proposed approach, we experiment with two Java applications. The experimental results showed that the proposed logistic regression model can accurately detect similar versions of software after learning from training data. The proposed logistic regression model is expected to be applied in applications for comparing and detecting similar versions of software.
Generally, Artificial Intelligence (AI) algorithms are unable to account for the logic of each decision they take during the course of arriving at a solution. This “black box” problem limits the usefulness of AI in military, medical, and financial security applications, among others, where the price for a mistake is great and the decision-maker must be able to monitor and understand each step along the process. In our research, we focus on the application of Explainable AI for log anomaly detection systems of a different kind. In particular, we use the Shapley value approach from cooperative game theory to explain the outcome or solution of two anomaly-detection algorithms: Decision tree and DeepLog. Both algorithms come from the machine learning-based log analysis toolkit for the automated anomaly detection “Loglizer”. The novelty of our research is that by using the Shapley value and special coding techniques we managed to evaluate or explain the contribution of both a single event and a grouped sequence of events of the Log for the purposes of anomaly detection. We explain how each event and sequence of events influences the solution, or the result, of an anomaly detection system.
Deep neural network (DNN) has shown significant improvement in learning and generalizing different machine learning tasks over the years. But it comes with an expense of heavy computational power and memory requirements. We can see that machine learning applications are even running in portable devices like mobiles and embedded systems nowadays, which generally have limited resources regarding computational power and memory and thus can only run small machine learning models. However, smaller networks usually do not perform very well. In this paper, we have implemented a simple ensemble learning based knowledge distillation network to improve the accuracy of such small models. Our experimental results prove that the performance enhancement of smaller models can be achieved through distilling knowledge from a combination of small models rather than using a cumbersome model for the knowledge transfer. Besides, the ensemble knowledge distillation network is simpler, time-efficient, and easy to implement.
Social media research has grown exponentially in recent years. However, it seems that to date only a few studies have applied Conversation Analysis (CA) to study social media interactions. The aim of this paper is to show the benefits of using CA for the analysis of this type of social interactions.
The neural network is an approach of machine learning by training the connected nodes of a model to predict the results of specific problems. The prediction model is trained by using previously collected training data. In training neural network models, overfitting problems can occur from the excessively dependent training of data and the structural problems of the models. In this paper, we analyze the effect of DropConnect for controlling overfitting in neural networks. It is analyzed according to the DropConnect rates and the number of nodes in designing neural networks. The analysis results of this study help to understand the effect of DropConnect in neural networks. To design an effective neural network model, the DropConnect can be applied with appropriate parameters from the understanding of the effect of the DropConnect in neural network models.
This paper summarizes the development and evolution from digital mock-up (DMU) to digital twin (DT) by clarifying the connotation of DT from prospective of digital product definition (DPD). Firstly, taking Airbus as example, the evolution of DMU is introduced, with the detailed analysis of configured DMU, functional DMU, and industrial DMU. Secondly, based on the literature review of DT, the definition, purpose and several applications of DT concept are clearly expounded. Finally, the augmentation for DT and DPD’s relationship are deduced.
The paper was trying to extract entities from related tweets collected from twitter. This project first collected real-time tweets from twitter searching API with related topic-based hashtags during the death of American black man George Floyd. We then used two approaches to identify the polarities or emotions of each tweets and generated over-time sentiment flow chart in detecting entities. We found that some extreme sentiment score was correlated with some key entities over time. And our adapted NRC-lexicon based approach obtained better results. This paper revealed that public’s sentiment displayed on tweets was generally consistent with the correlated events previously. It might help researchers in predicting or preventing public events in the future.