Search

Nowadays, the improvement of digital learning with Artificial Intelligence has attracted a lot of research, as it provides solutions for individualized education styles which are independent of place and time. This is particularly the case for computer science, as a tutoring domain, which is rapidly growing and changing and as such, learners need frequent update courses. In this paper, we present a thorough evaluation of a fuzzy-based intelligent tutoring system (ITS), that teaches computer programming. The evaluation concerns multiple aspects of the ITS. The evaluation criteria are: (i) context, (ii) effectiveness, (iii) efficiency, (iv) accuracy, (v) usability and satisfaction, and (vi) engagement and motivation. In the evaluation process students of an undergraduate program in Informatics of the University of Piraeus in Greece participated. The evaluation method that was used included questionnaires, analysis of log files and experiments. Also, t-tests were conducted to certify the validity of the evaluation results. Indeed, the evaluation results are very positive and show that the incorporated fuzzy mechanism to the presented ITS enhances the system with Artificial Intelligence and through this, it increases the learners' satisfaction and new knowledge learning and mastering, improves the recommendation accuracy of the system, the efficacy of interactions, and contributes positively to the learners' engagement in the learning process.

In this paper, we present a novel Artificial Intelligence (AI) -empowered system that enhances large language models and other machine learning tools with rules to provide primary care diagnostic advice to patients. Specifically, we introduce a novel methodology, represented through a process diagram, which allows the definition of generative AI processes and functions with a focus on the rule-augmented approach. Our methodology separates various components of the generative AI process as blocks that can be used to generate an implementation data flow diagram. Building upon this framework, we utilize the concept of a dialogue process as a theoretical foundation. This is specifically applied to the interactions between a user and an AI-empowered software program, which is called "Med|Primary AI assistant" (Alpha Version at the time of writing), and provides symptom analysis and medical advice in the form of suggested diagnostics. By leveraging current advancements in natural language processing, a novel approach is proposed to define a blueprint of domain-specific knowledge and a context for instantiated advice generation. Our approach not only encompasses the interaction domain, but it also delves into specific content that is relevant to the user, offering a tailored and effective AI–user interaction experience within a medical context. Lastly, using an evaluation process based on rules, defined by context and dialogue theory, we outline an algorithmic approach to measure content and responses. [ABSTRACT FROM AUTHOR]

The Intelligent Decision Technologies (IDT) Journal is a scholarly publication that focuses on the theory and application of intelligent technologies and systems for decision making. It was founded in 2007 by Professors Gloria Phillips-Wren and Lakhmi C. Jain to address the need for real-world decision support systems. The journal has gained popularity over the years and now receives over 800 submissions annually. It covers various topics related to intelligent decision support, such as artificial intelligence and data processing. The current editors-in-chief are Professors George A. Tsihrintzis, Gloria Phillips-Wren, and Lakhmi C. Jain. The text provides information about the journal's recognition, the number of submitted and published papers, special collections and issues, and future plans. It also acknowledges the contributions of associate editors and encourages readers to explore the journal's website for more information and to consider submitting their research work. The editors hope that readers will find the journal useful and inspiring. [Extracted from the article]

This document is an editorial note introducing a special issue of the journal Electronics on the topic of "Trends and Applications in Information Systems and Technologies." The note highlights the importance of information technologies in reshaping business models and improving operational efficiency. It also discusses the challenges associated with information systems, such as cybersecurity issues and data-centric decision-making processes. The special issue includes fourteen selected papers covering topics such as the adoption of blockchain technology, digital business solutions, ICT penetration, processing and information extraction in information systems, and innovative applications of information systems. The note concludes by expressing the hope that readers will find the special issue informative and inspiring for further research. [Extracted from the article]

This paper presents a novel development methodology for artificial intelligence (AI) analytics in energy management that focuses on tailored explainability to overcome the "black box" issue associated with AI analytics. Our approach addresses the fact that any given analytic service is to be used by different stakeholders, with different backgrounds, preferences, abilities, skills, and goals. Our methodology is aligned with the explainable artificial intelligence (XAI) paradigm and aims to enhance the interpretability of AI-empowered decision support systems (DSSs). Specifically, a clustering-based approach is adopted to customize the depth of explainability based on the specific needs of different user groups. This approach improves the accuracy and effectiveness of energy management analytics while promoting transparency and trust in the decision-making process. The methodology is structured around an iterative development lifecycle for an intelligent decision support system and includes several steps, such as stakeholder identification, an empirical study on usability and explainability, user clustering analysis, and the implementation of an XAI framework. The XAI framework comprises XAI clusters and local and global XAI, which facilitate higher adoption rates of the AI system and ensure responsible and safe deployment. The methodology is tested on a stacked neural network for an analytics service, which estimates energy savings from renovations, and aims to increase adoption rates and benefit the circular economy. [ABSTRACT FROM AUTHOR]

In this paper, we present an Intelligent Tutoring System (ITS), for use in teaching the logic of computer programming and the programming language 'C'. The aim of the ITS is to adapt the delivered learning material and the lesson sequence to the knowledge level and learning needs of each individual student. The adaptation of the presented ITS is based on fuzzy logic and a machine learning technique. Particularly, the system uses the distance weighted k-nearest neighbor algorithm to detect the learner's knowledge level and abilities concerning computer programming during her/ his first interaction with the system. Next and during subsequent interactions of the learner with the system, fuzzy logic is used to identify the learner's current knowledge level and potential misconceptions. The system takes into consideration the knowledge dependencies that exist among the domain concepts of the learning material and, applying fuzzy rules, decides about the learning material that has to be delivered to the learner as well as the lesson sequence. The system has been fully implemented and evaluated through t-tests. The evaluation results show that the combination of machine learning (for initially identifying the student's learning abilities and needs) with fuzzy logic (for the continuous identification of the learner's current knowledge level and misconceptions) provides more personalized learning experience, promotes the active participation of students in the learning process and results in decrease in the number of dropouts. [ABSTRACT FROM AUTHOR]

In this paper, a fuzzy-based evaluation method is presented for the impact of e-learning on several aspects of the learning lives of academic students of Information and Communication Technologies (ICT) during the COVID-19 pandemic. Specifically, the academic year 2020–2021 was considered when a general lockdown was imposed in Greece and all courses were delivered exclusively through the web. The evaluation criteria are (i) e-learning acceptance, (ii) learning effectiveness, (iii) students' engagement and (iv) socializing and interpersonal relationships in the educational community. The evaluation was conducted through questionnaires. Ninety two (92) undergraduate and postgraduate students and Ten (10) computer science in-structors of the department of Informatics of the University of Piraeus, Greece, participated in this survey. The questionnaire answers were analyzed using a fuzzy-based mechanism. Particularly, five fuzzy sets were used to describe the 5-point Likert scale answers to the questionnaires with linguistic values, and four other fuzzy sets were used for the description of the evaluation results concerning the four defined evaluation criteria. Moreover, 209 fuzzy rules were created to analyze and calculate the evaluation results per criterion, considering both the students' answers (119 fuzzy rules) and the instructors' answers (90 fuzzy rules) to the questionnaires. The gain of this approach is that the analysis of evaluation data with fuzzy rules imitates the human way of thinking and, thus, makes the process more explainable. The evaluation results showed a positive impact of e-learning on student confidence, self-discipline and active participation and a negative impact on student interpersonal relationships. The impact of e-learning on other learning issues was neutral. [ABSTRACT FROM AUTHOR]

Artificial Intelligence research is presenting phenomenal progress in two directions: (i) new theories and methodologies, and (ii) applications that expand traditional domains with innovative interventions. As indicated by recent reports, this progress has created a disequilibrium, where demand for scientists with skills in Artificial Intelligence is not fulfilled, a trend that will intensify further in the years to come. A potential solution to this shortage of specialised workforce may come from encouraging more women to get educated and follow a career in one of the Artificial Intelligence areas. This special collection of invited papers is dedicated to all women researchers and practitioners in Artificial Intelligence and coincides with the March 8, 2023 International Women's Day. Moreover, it has two specific goals: (i) to inspire more women to study and practice Artificial Intelligence through presentation of recognized women researchers who can act as role models, and (ii) to highlight some streamlined research areas of Artificial Intelligence. [ABSTRACT FROM AUTHOR]

This paper presents the most recent version of V-GRAFFER, a novel system that we have been developing for Visual GRoup AFFEct Recognition research. This version includes new algorithms and features, as well as a new application extension for using and evaluating the new features. Specifically, we present novel methods to collect facial samples from other e-lecture applications. We use screen captures of lectures, which we track and connect with samples during the duration of e-educational events. We also developed and evaluated three new algorithms for drawing conclusions on group concentration states. As V-GRAFFER required such complex functionalities to be combined together, many corresponding microservices have been developed. The current version of V-GRAFFER allows drawing real-time conclusions using the input samples collected from the use of any tutoring system, which in turn leads to real-time feedback and allows adjustment of the course material. [ABSTRACT FROM AUTHOR]

Testing is a significant part of the teaching and learning process. An assessment test has to include test items that are tailored to the individual learning needs of the students in order to be more accurate and support learning in a more effective way. In this paper, a fuzzy-based mechanism is presented for automatic personalized assessment in an e-learning system for computer programming. Particularly, the selection of the most appropriate test items for each individual student is based on a variety of criteria: (i) the student's knowledge level, (ii) the student's prior knowledge of computer programming, (iii) the type of programming errors that the student is prone to make, and (iv) the difficulty level of the test items. Linguistic values are used to determine these criteria. Additionally, 45 fuzzy rules are used over these criteria, which imitate the way of thinking of human tutors with regard to deciding about the most appropriate test items that have to be included in an adaptive test. The presented mechanism was used under real conditions and evaluated by experts and students of the Department of Informatics of the University of Piraeus, Greece with very encouraging results. Specifically, both the participating students and experts found that the presented mechanism creates non-repetitive balanced tests that meet learners' knowledge level and needs. [ABSTRACT FROM AUTHOR]

In this paper, we present our research towards building V-GRAFFER, a system for Visual GRoup AFFect Recognition. Specifically, V-GRAFFER aims at the development and provision of services with regard to the recognition of the emotional state of groups of people. At this stage of the V-GRAFFER development, implemented services are oriented towards detecting and drawing conclusions about students who attend educational events such as lectures, question and answer (Q&A) sessions, or lab participation. Specific functionalities of the current version of V-GRAFFER include processes for data collection, lecture experiments under real conditions, algorithms for sample auto-extraction, and optimized approaches. These functionalities allowed the collection of data of various educational events under real conditions and the creation of flexible databases of appropriate samples for drawing conclusions with regard to emotion detection of groups of people. Furthermore, we devised and implemented innovative algorithms to identify and classify group samples from recorded educational events. These algorithms have been evaluated and improved via continuous cycles of development-testing-evaluation in a variety of experiments. Finally, we constructed completed databases of group samples which are correlated with each other based on time, depth of time and educational settings. [ABSTRACT FROM AUTHOR]

Biomarkers have already been proposed as powerful classification features for use in the training of neural network-based and other machine learning and artificial intelligence-based prognostic models in the scientific field of personalized nutrition. In this paper, we construct and study cascaded SVM-based classifiers for automated metabolic syndrome diagnosis. Specifically, using blood exams, we achieve an average accuracy of about 84% in correctly classifying body mass index. Similarly, cascaded SVM-based classifiers achieve a 74% accuracy in correctly classifying systolic blood pressure. Next, we propose and implement a system that achieves an 84% accuracy in metabolic syndrome prediction. The proposed system relies not only on prediction of the body mass index but also on prediction from blood exams of total cholesterol, triglycerides and glucose. For the aim of self-completeness of the paper, the key concepts with regard to metabolic syndrome are summarized, and a review of previous related work is included. Finally, conclusions are drawn and indications for related future research are outlined. [ABSTRACT FROM AUTHOR]

Many humans deal with problems that concern episodic memory decline. These problems can cause annoying, and sometimes, dangerous incidents, like failing to recall the name of a friend or forgetting to take a medicine or turn off the cooker. Concerning the above, this paper presents a fuzzy rule-based mechanism that generates emergency alerts when a dangerous situation is caused by an inconsistency in the human's actions in the home environment. In such a way, the system protects persons with episodic memory decline problems or lapses of attention from dangerous situations that may be caused by their memory disorder and allows them to complete an everyday activity. For the paper's needs the application of the presented mechanism is limited to the usage of a cooker. The system takes as input the time, the size and the content of the pot that is used to the cooker, calculates the degree of emergency and describes it using fuzzy sets and, finally, applies rules over the fuzzy sets to generate alert messages that notify the monitored person about the next action that s/he has to do in order to complete a particular activity. For the evaluation of the system, we have developed a simulation program that asks users to complete some activities during a specific time period. The system embeds the presented fuzzy rule-based mechanism and monitors the user's actions and generated alerts, which concern the usage of the stove burner. The simulation software was used by 15 users. Their reactions and opinions about the system's alerts and the assistance it offers, are positive. [ABSTRACT FROM AUTHOR]

The doctrine of the "one size fits all" approach in the field of disease diagnosis and patient management is being replaced by a more per patient approach known as "personalized medicine". In this spirit, biomarkers are key variables in the research and development of new methods for prognostic and classification model training based on advances in the field of artificial intelligence [1, 2, 3]. Metabolomics refers to the systematic study of the unique chemical fingerprints that cellular processes leave behind. The metabolic profile of a person can provide a snapshot of cell physiology and, by extension, metabolomics provide a direct "functional reading of the physiological state" of an organism. Via employing machine learning methodologies, a general evaluation chart of nutritional biomarkers is formulated and an optimised prediction method for body to mass index is investigated with the aim to discover dietary patterns. [ABSTRACT FROM AUTHOR]

In emergency situations, different actors involved in first aid services should be authorized to retrieve information from the patient's Electronic Health Records (EHRs). The research objectives of this work involve the development and implementation of methods to characterise emergency situations requiring extraordinary access to healthcare data. The aim is to implement such methods based on contextual information pertaining to specific patients and emergency situations and also leveraging personalisation aspects which enable the efficient access control on sensitive data during emergencies. The Attribute Based Access Control paradigm is used in order to grant access to EHRs based on contextual information. We introduce an ABAC approach using personalized context handlers, in which raw contextual information can be uplifted in order to recognize critical situations and grant access to healthcare data. Results indicate that context-aware ABAC is a very effective method for detecting critical situations that require emergency access to personal health records. In comparison to RBAC implementations of emergency access control to EHRs, the proposed ABAC implementation leverages contextual information pertaining to the specific patient and emergency situations. Contextual information increases the capability of ABAC to recognize critical situations and grant access to healthcare data. [ABSTRACT FROM AUTHOR]

Addressing undeclared work is a high priority in the labor field for government policymakers since it adversely affects all involved parties and results in significant losses in tax and social security contribution revenues. In the last years, the wide use of ICT in labor inspectorates and the considerable progress in data exchange have resulted in numerous databases dispersed in various units, yet these are not effectively used to increase their functions productivity. This study presents a detailed analysis of a data mining project per the CRISP-DM methodology aiming to assist the labor inspectorates in dealing with undeclared work and other labor law violations. It uses real past inspections data merged with companies characteristics and their employment details and examines the application of two Associative Classification algorithms, the CBA and CBA2, in combination with two types of datasets, a binary and a four-class. The produced models are assessed per the data mining goals and per the initial business objectives, and the research concludes proposing an innovative inspections recommendation tool proved to offer two major benefits: a mechanism for planning targeted inspections of improved efficiency and a knowledge repository for enhancing the inspectors understanding of those features linked with labor law violations. [ABSTRACT FROM AUTHOR]

Neuro-symbolic approaches combine neural and symbolic methods. This paper explores aspects regarding the reasoning mechanisms of two neuro-symbolic approaches, that is, neurules and connectionist expert systems. Both provide reasoning and explanation facilities. Neurules are a type of neuro-symbolic rules tightly integrating the neural and symbolic components, giving pre-eminence to the symbolic component. Connectionist expert systems give pre-eminence to the connectionist component. This paper explores reasoning aspects about neurules and connectionist expert systems that have not been previously addressed. As far as neurules are concerned, an aspect playing a role in conflict resolution (i.e., order of neurules) is explored. Experimental results show an improvement in reasoning efficiency. As far as connectionist expert systems are concerned, variations of the reasoning mechanism are explored. Experimental results are presented for them as well showing that one of the variations generally performs better than the others. [ABSTRACT FROM AUTHOR]

Energy communities can support the energy transition, by engaging citizens through collective energy actions and generate positive economic, social and environmental outcomes. Renewable Energy Sources (RES) are gaining increasing share in the electricity mix as the economy decarbonises, with Photovoltaic (PV) plants to becoming more efficient and affordable. By incorporating Artificial Intelligence (AI) techniques, innovative applications can be developed to provide added value to energy communities. In this context, the scope of this paper is to compare Machine Learning (ML) and Deep Learning (DL) algorithms for the prediction of short-term production in a solar plant under an energy cooperative operation. Three different cases are considered, based on the data used as inputs for forecasting purposes. Lagged inputs are used to assess the historical data needed, and the algorithms' accuracy is tested for the next hour's PV production forecast. The comparative analysis between the proposed algorithms demonstrates the most accurate algorithm in each case, depending on the available data. For the highest performing algorithm, its performance accuracy in further forecasting horizons (3 hours, 6 hours and 24 hours) is also tested. [ABSTRACT FROM AUTHOR]

Animal Vehicle Collision (AVC) is relatively an evolving source of fatality resulting in the deficit of wildlife conservancy along with carnage. It's a globally distressing and disturbing experience that causes monetary damage, injury, and human-animal mortality. Roadkill has always been atop the research domain and serendipitously provided heterogeneous solutions for collision mitigation and prevention. Despite the abundant solution availability, this research throws a new spotlight on wildlife-vehicle collision mitigation using highly efficient artificial intelligence during nighttime hours. This study focuses mainly on arousal mechanisms of the "Histogram of Oriented Gradients (HOG)" intelligent system with extracted thermography image features, which are then processed by a trained, convolutional neural network (1D-CNN). The above computer vision – deep learning-based alert system has an accuracy between 94%, and 96% on the arousal mechanisms with the empowered real-time data set utilization. [ABSTRACT FROM AUTHOR]

Reliable and rapid non-invasive testing has become essential for COVID-19 diagnosis and tracking statistics. Recent studies motivate the use of modern machine learning (ML) and deep learning (DL) tools that utilize features of coughing sounds for COVID-19 diagnosis. In this paper, we describe system designs that we developed for COVID-19 cough detection with the long-term objective of embedding them in a testing device. More specifically, we use log-mel spectrogram features extracted from the coughing audio signal and design a series of customized deep learning algorithms to develop fast and automated diagnosis tools for COVID-19 detection. We first explore the use of a deep neural network with fully connected layers. Additionally, we investigate prospects of efficient implementation by examining the impact on the detection performance by pruning the fully connected neural network based on the Lottery Ticket Hypothesis (LTH) optimization process. In general, pruned neural networks have been shown to provide similar performance gains to that of unpruned networks with reduced computational complexity in a variety of signal processing applications. Finally, we investigate the use of convolutional neural network architectures and in particular the VGG-13 architecture which we tune specifically for this application. Our results show that a unique ensembling of the VGG-13 architecture trained using a combination of binary cross entropy and focal losses with data augmentation significantly outperforms the fully connected networks and other recently proposed baselines on the DiCOVA 2021 COVID-19 cough audio dataset. Our customized VGG-13 model achieves an average validation AUROC of 82.23% and a test AUROC of 78.3% at a sensitivity of 80.49%. [ABSTRACT FROM AUTHOR]

The new user cold-start problem is a grand challenge in content-based music recommender systems. This happens when the systems do not have sufficient information regarding the user's preferences. Towards solving this problem, in this study, a rating prediction framework is proposed. The proposed framework allows the systems to predict the user's rating scores for unrated musical pieces, by which good recommendations can be generated. The core idea here is to leverage the so-called MUSIC model, i.e., a five-factor musical preference model, which is characterized by Mellow, Unpretentious, Sophisticated, Intense, and Contemporary as the user's musical preference profiles. When a user newly joins the systems, the first five-factor musical preference profile is established based on the user's age and brain type information which is extracted from questionnaires. When the user experiences the systems for a certain period, his/her rating scores for experienced musical pieces are utilized for generating the second five-factor musical preference profile. The recommendations are then provided based on the rating scores predicted from a non-linear combination of these two five-factor musical preference profiles. The results demonstrated the effectiveness of the five-factor musical preference in alleviating the new user cold-start problem. In addition, the proposed method can potentially provide high-quality recommendations. [ABSTRACT FROM AUTHOR]

We introduce the DP-auto-GAN framework for synthetic data generation, which combines the low dimensional representation of autoencoders with the flexibility of Generative Adversarial Networks (GANs). This framework can be used to take in raw sensitive data and privately train a model for generating synthetic data that will satisfy similar statistical properties as the original data. This learned model can generate an arbitrary amount of synthetic data, which can then be freely shared due to the post-processing guarantee of differential privacy. Our framework is applicable to unlabeled mixed-type data, that may include binary, categorical, and real-valued data. We implement this framework on both binary data (MIMIC-III) and mixed-type data (ADULT), and compare its performance with existing private algorithms on metrics in unsupervised settings. We also introduce a new quantitative metric able to detect diversity, or lack thereof, of synthetic data. [ABSTRACT FROM AUTHOR]

This paper is aimed at helping organizations to understand what they can expect from a serverless architecture in the future and how they can make sound decisions about the choice between microservice and serverless architectures in the present. A serverless architecture is a new approach to offering services in the cloud. It was invented as a solution to the problem that many organizations are facing today – about 85% of their servers have underutilized capacity, which is proved to be costly and wasteful. By employing the serverless architecture, the organizations get a way to eliminate idle, underutilized servers and thus, to reduce their operational costs. Many cloud providers are now jumping to the serverless world because they know it is going to be the future of software architectures. However, being a new approach, the serverless architecture is still relatively immature – it is in the early stages of its support by cloud service platform providers. This paper provides an in-depth study about the serverless architecture and how to apply FaaS in the real world. [ABSTRACT FROM AUTHOR]

In short, clustering is the process of partitioning a given set of objects into groups containing highly related instances. This relation is determined by a specific distance metric with which the intra-cluster similarity is estimated. Finding an optimal number of such partitions is usually the key step in the entire process, yet a rather difficult one. Selecting an unsuitable number of clusters might lead to incorrect conclusions and, consequently, to wrong decisions: the term "optimal" is quite ambiguous. Furthermore, various inherent characteristics of the datasets, such as clusters that overlap or clusters containing subclusters, will most often increase the level of difficulty of the task. Thus, the methods used to detect similarities and the parameter selection of the partition algorithm have a major impact on the quality of the groups and the identification of their optimal number. Given that each dataset constitutes a rather distinct case, validity indices are indicators introduced to address the problem of selecting such an optimal number of clusters. In this work, an extensive set of well-known validity indices, based on the approach of the so-called relative criteria, are examined comparatively. A total of 26 cluster validation measures were investigated in two distinct case studies: one in real-world and one in artificially generated data. To ensure a certain degree of difficulty, both real-world and generated data were selected to exhibit variations and inhomogeneity. Each of the indices is being deployed under the schemes of 9 different clustering methods, which incorporate 5 different distance metrics. All results are presented in various explanatory forms. [ABSTRACT FROM AUTHOR]

B Information b is widely available and accessible, but frequently leads to information overload and overexposure, while the effort for coding, storing, hiding, securing, transmitting and retrieving it may be excessive. Special Collection of Extended Selected Papers on "Novel Research Results Presented in The 12th International Conference on Information, Intelligence, Systems and Applications (IISA2021), 12-14 July 2021, Chania, Crete, Greecehttps://easyconferences.eu/iisa2021/" The authors employ machine learning-based methodologies to formulate a general evaluation chart of nutritional biomarkers. [Extracted from the article]

Biomarkers are measurements of biological variables that can determine a state of health. They consist of measuring a single variable or a combination of variables related to the state of health that these variables represent. Biomarkers can provide an early warning of a health problem in relation to an individual patient or group of patients, and thus trigger actions and lead to interventions. Nutritional biomarkers measure the biological consequences of one's diet. In our recent work, we have used machine learning to predict weight, metabolic syndrome and blood pressure, using blood-exam-based biomarkers. In the current work, we use extreme value theory to examine the significance of outliers in health data, with a focus on diet and the standard biochemistry profile. Specifically, we show that, using extreme value analysis and by applying a systemisation of the process, health trends can be predicted, and thus, health interventions can be (at least partially) automated. For that purpose, public access datasets have been used, which were retrieved from the National Health and Nutrition Examination Survey. The NHANES is a program of studies designed to assess the health and nutritional status of the population in the United States. In total, about 70,000 datapoints were analysed, covering about a decade's worth of observations. [ABSTRACT FROM AUTHOR]

In this paper, we present a handwritten character recognition (HCR) system that aims to recognize first-order logic handwritten formulas and create editable text files of the recognized formulas. Dense feedforward neural networks (NNs) are utilized, and their performance is examined under various training conditions and methods. More specifically, after three training algorithms (backpropagation, resilient propagation and stochastic gradient descent) had been tested, we created and trained an NN with the stochastic gradient descent algorithm, optimized by the Adam update rule, which was proved to be the best, using a trainset of 16,750 handwritten image samples of 28 × 28 each and a testset of 7947 samples. The final accuracy achieved is 90.13%. The general methodology followed consists of two stages: the image processing and the NN design and training. Finally, an application has been created that implements the methodology and automatically recognizes handwritten logic formulas. An interesting feature of the application is that it allows for creating new, user-oriented training sets and parameter settings, and thus new NN models. [ABSTRACT FROM AUTHOR]