IDEAL2021: Special Sessions

---

Scope

Clustering is one of the most important tasks in Data Analysis, Data Mining, and Knowledge Discovery. Clustering applications covers data exploration, hypothesis generating and/or testing, model fitting, and typology finding. The societal requirements on interpretable Artificial Intelligence (AI) systems leads to considerable effort to develop interpretable models in Machine Learning and approaches to explain the decision/prediction processes to decision-makers. This raises exciting challenges to design more effective clustering methods and tools able to extract valuable information from data, imposing new questions for clustering interpretability. Data Recovery approach to Clustering, also referred to as the Auto-Encoder Clustering in the Neural Networks community, is in the realm of modern Data Analysis. The coupled encode-decode Data Recovery approach to Clustering provides a flexible data-driven model estimation with rich generalization and interpretable capabilities. 

The aim of this session is to offer a forum for both academic and industrial communities to share and disseminate innovative research efforts and developments to modern Data Analysis for interpretable AI. Topics of interest include but are not limited to:

 • Unsupervised/Semi-supervised (fuzzy) clustering

 • Unsupervised/Semi-supervised feature extraction

 • Unsupervised/Semi-supervised anomaly/novelty/outlier detection

 • Unsupervised/Semi-supervised stream analysis and trend detection

 • Neurocomputing, fuzzy neural nets, and deep learning 

 • Deep clustering

 • Methods for interpretation of clusters

 • Fuzzy evolutionary computation

 • Data Visualization

 • Methods for interpretation of clusters

 • Methods to deal with uncertainty and interpretability in data processing

 • Applications of unsupervised/semi-supervised clustering

 • Artificial Intelligence for uncertain environments

Organisers

Susana Nascimento, NOVA University, Lisboa, Portugal. Email: snt@fct.unl.pt

José Valente de Oliveira, University of Algarve, Faro, Portugal. Email: jvo@ualg.pt

Victor Sousa Lobo, NOVA University, Lisboa, Portugal. Email: vlobo@novaims.unl.pt

Boris Mirkin, National Research University, HSE, Moscow, Russian Federation. Email: bmirkin@hse.ru

Scope

Multimodal Systems research is focused in methods and tools “to create, access, and interact all forms of digital content in any device or scenario”. Multimodal human-computer interaction is related to the user interaction with the virtual and physical scenario through the most natural modes of communication. Nowadays, with the advent of Ubiquitous Computing (Ubicomp), which focuses on a human-centred paradigm, multimodal systems aim to provide interaction with adaptive content, services, and interfaces towards each one of its users, according to the context of the applications’ scenarios. However, the provision of that appropriated personalised interaction is a true challenge due to different reasons, such as the user interests, heterogeneous environments and devices, or the dynamic user behaviour and data capture. 

Machine Learning (ML) algorithms analyse historical data, build models & predict outcomes. Machine Learning has got vast industry transformation capabilities and is used across sectors for business efficiency. Applications that provide large amounts of data regarding the users' interaction, such as usage data, preferences or even context of usage, should use machine learning to reason about the users. It should be used to acquire models of individual users interacting with the information system and grouping them into communities or stereotypes with common interests. In this regard, ML techniques can be excellent tools to cope with difficult problems that arise when implementing smarter multimodal systems, ranging from representation, translation, alignment or fusion between modalities, especially for Ubicomp scenarios. There are many applications which can be tackled by ML techniques, such as the personalization of the user interaction, the distribution of the user interface across different devices, or new user-interface approaches, towards intelligent user interactions with multimodal systems to better reach the Ubicomp vision. 

This special session aims to cover a wide range of works and recent advances on the application of ML techniques to enhance the user interaction in Multimodal Systems. We hope that this session can provide a common forum for researchers and practitioners to exchange their ideas and report their latest findings in the following topics. 

.

• Intelligent User Interaction.

• Personalization of the User Interaction.

• Recommendation in multimodal systems.

• Modelling human behavior.

• Machine Learning for Accessibility.

• Prediction of future users’ activities and interactions.

• Distribution of the User interface across different devices.

• Novel user-interface techniques.

• Machine Learning for healthcare applications.

• Machine Learning for Creativity.

• Machine Learning towards Ubicomp systems.

• Any application of ML techniques to HCI problems.

Organisers

Nuno Correia, NOVA LINCS, DI, FCT, NOVA University of Lisboa, Portugal. Email: nmc@fct.unl.pt

Rui Neves Madeira, NOVA LINCS, DI, FCT, NOVA University of Lisboa and Sustain.RD center, Polytechnic Inst. Setúbal, Portugal. Email: rui.madeira@estsetubal.ips.pt

Susana Nascimento, NOVA LINCS, DI, FCT, NOVA University of Lisboa, Portugal. Email: snt@fct.unl.pt 

Scope

Computer vision and image processing is a big research field that studies

to use computers to process, extract, analyze and understand information from

digital images and videos as the human vision system does. It covers a wide

range of applications in many important fields, including engineering, biology,

medicine, remote sensing, and business. Furthermore, many computer vision

tasks are highly related to our daily life, from face detection in the mobile

phone to self-driving vehicles. Typical tasks related to computer vision and

image processing include edge detection, image classification, image

segmentation, object detection, object recognition, scene analysis, biological

identification, motion analysis, image restoration, and image enhancement.

These tasks have not been comprehensively solved, particularly in the era of

big data, when image data are easy to obtain but may be more challenging to

analyze. It is necessary to develop new effective and efficient methods to solve

these tasks.

Computational intelligence (CI) is a sub-field of artificial intelligence that

includes a big family of biologically and linguistically motivated

computational paradigms, including neural networks, evolutionary

computation and fuzzy systems. CI plays a major role in developing successful

intelligent systems for computer vision and image processing. In the last

decades, CI techniques have been widely applied to computer vision and

image processing and achieved promising results. However, the potential of CI

has not been fully explored. The challenges of improving effectiveness,

efficiency, and interpretability, and reducing the requirement of domain

knowledge are urging to be further addressed by investigating new

computational paradigms to computer vision and image analysis.

This special session aims to investigate the use of computational

intelligence for computer vision and image analysis, covering ALL different

computation paradigms and their applications to computer vision and image

analysis. It will bring together researchers and practitioners from around the

world to discuss the latest advances in the field and will act as a major forum

for the presentation of recent research. Authors are invited to submit their

original and unpublished work to this special session. Topics related to all

aspects of computational intelligence for computer vision and image analysis,

such as theories, algorithms, systems and applications, are welcome.

Topics of interest include but are not limited to: 

▪ Neural networks

▪ Convolutional neural networks (CNNs)

▪ Autoencoder

▪ Recurrent neural networks (RNNs)

▪ Neural architecture search (NAS)

▪ Deep neural networks (DNNs)

▪ Recurrent neural networks (RNNs)

▪ Long short-term memory (LSTMs)

▪ Deep belief networks (DBNs)

▪ Generative adversarial networks (GANs)

▪ Other neural network variants

▪ Evolutionary computation

▪ Evolutionary algorithms (EAs)

▪ Swarm intelligence (SI)

▪ Evolutionary multi-objective optimization (EMO)

▪ Genetic algorithms (GAs)

▪ Genetic programming (GP)

▪ Particle swarm optimisation (PSO)

▪ Ant colony optimisation (ACO)

▪ Differential evolution (DE)

▪ Evolutionary transfer learning/optimisation (ETO)

▪ Evolutionary multitask learning/optimisation

▪ Others

▪ Fuzzy system

▪ Fuzzy logic

▪ Fuzzy sets

▪ Fuzzy control

▪ Granular computing

▪ Rough set

▪ Hybridisations of different CI techniques

▪ Meta learning

▪ Transfer learning

▪ Other bio-inspired computation paradigms

and their applications to

▪ Image processing

▪ Image classification

▪ Image segmentation

▪ Semantic segmentation

▪ Instance segementation

▪ Image enhancement

▪ Edge detection

▪ Image restoration

▪ Image registration

▪ Object detection

▪ Object recognition

▪ Object tracking

▪ Image retrieval

▪ Texture analysis

▪ Scene analysis and understanding

▪ Face recognition

▪ Facial expression and emotion analysis

▪ Biological identification

▪ Action recognition and human activity analysis

▪ Medical image analysis

▪ Remote sensing image analysis

▪ Feature extraction and analysis

▪ Representation learning

▪ Pattern recognition

▪ Big data in computer vision and image analysis

▪ Small data in computer vision and image analysis

▪ Other computer vision and image analysis applications

Organisers

Ying Bi, School of Engineering and Computer Science, Victoria University of

Wellington, New Zealand. Email: Ying.Bi@ecs.vuw.ac.nz

Bing Xue, School of Engineering and Computer Science, Victoria University of

Wellington, New Zealand. Email: Bing.Xue@ecs.vuw.ac.nz

Antonio J. Tallón-Ballesteros, University of Huelva, Spain. Email: antonio.tallon.diesia@zimbra.uhu.es

Scope

Many modern applications in data engineering and machine learning demand huge computational resources and require substantial amounts of energy. Reconfigurable hardware technology allows for the implementation of application-specific accelerators that can achieve great improvements in performance and energy efficiency through customized operators and interconnects, applications-specific wide data paths and memory architectures, as well as massively parallel execution and deep pipelines.

The aim of this session is to bring together researchers and engineers working on reconfigurable accelerators for data engineering and machine learning with the goal to discuss and exchange latest ideas and results. The topics of interest for this session include, but are not limited to:

• The use of FPGAs for Deep Learning

• FPGA-based Machine Learning Accelerators

• Reconfigurable Architectures for Machine Learning

• Reconfigurable Architectures for Data Mining and Data Analytics

• Reconfigurable Architectures for Information Retrieval

• Bio- And Neuro-inspired Reconfigurable Computing

• Programming and Compiling Machine Learning techniques to FPGAs

• Reconfigurable Computing for Intelligent Techniques and AI

• Use of Machine Learning in FPGA Programming and Design Tasks

• Energy-Efficient Reconfigurable Computing for Intelligent Systems

We are interested in accelerated data engineering and machine learning at all levels of the performance spectrum, from HPC/data center down to edge/IoT systems.

Organisers

Marco Platzner, Paderborn University, Germany. Email: platzner@upb.de

João MP Cardoso, University of Porto, Portugal. Email: jmpc@computer.org 

Antonio J. Tallón-Ballesteros, University of Huelva, Spain. Email: antonio.tallon.diesia@zimbra.uhu.es

Scope and Aim

Imbalanced classification is one of the most important tasks in machine learning, which has attracted much attention from both academic and industrial communities. Imbalanced classification has a very wide range of real-world applications, most of which are closely related to our daily life, such as medical diagnosis, intrusion detection, anomaly detection, and credit card fraud detection. Imbalanced data exhibits a skewed distribution between its classes. If the class imbalance issue is not well-addressed, classifiers are likely to ignore the class of interest which is constituted by a few instances. Computational intelligence is a subfield of artificial intelligence, which covers three research branches, including evolutionary computation, fuzzy sets, and artificial neural networks. Computational intelligence techniques have been applied and achieved great contributions to imbalanced classification. In the big data era, the amount of data is growing very rapidly, either increasing in a number of features or instances. This brings further difficulty in constructing effective classifiers when learning from imbalanced data but in return enriches new opportunities. 

The aim of this special session is to join the contemporary use of computational intelligence techniques for imbalanced classification. This special session attempts to bring together some of leading experts or researchers from different branches in computational intelligence and offers a forum for them to present their latest research, discuss open questions as well as current advances in imbalanced classification. The session welcomes studies and contributions that introduce novel methods based on different computational intelligence paradigms to imbalanced classification and its applications.

Authors are invited to submit their original and unpublished work to this special session, which is concerned with imbalanced data and approaches based on computational intelligence techniques under the umbrella of imbalanced classification. Topics of interest include but are not limited to:

• Evolutionary computation (e.g. Genetic Algorithms, Genetic Programming and Particle Swarm Optimization, etc.) for imbalanced classification.

• Evolutionary computation with sampling methods (including undersampling, oversampling and hybrid sampling) for imbalanced classification.

• Evolutionary computation with cost-sensitive learning for imbalanced classification.

• Fuzzy sets, Rough sets, Granular computing for imbalanced classification.

• Fuzzy rule-based classification systems for imbalanced classification.

• Neural networks for imbalanced classification.

• Deep learning for imbalanced classification.

• Sampling methods for imbalanced classification.

• Instance selection for large-scale imbalanced data.

• Cost-sensitive learning for imbalanced classification.

• Active learning for imbalanced classification.

• Improve interpretability of over-complicated models for imbalanced classification.

• Feature selection/construction/extraction/ranking/analysis for imbalanced classification with high-dimensional data.

• Real-world applications of imbalanced classification, e.g., medical data analysis, fault detection, anomaly detection, software defect prediction, and text mining.

Organisers

Wenbin Pei, Victoria University of Wellington, New Zealand. Email: Wenbin.Pei@ecs.vuw.ac.nz

Dr. Bing Xue, Victoria University of Wellington, New Zealand. Email: Bing.Xue@ecs.vuw.ac.nz

Antonio J. Tallón-Ballesteros, University of Huelva, Spain. Email: antonio.tallon.diesia@zimbra.uhu.es

Scope

Adaptive educational systems are technologically advanced applications aiming to provide immediate and personalized instruction or feedback to learners. Their extensive use in education is completely transforming the way we teach and learn. In addition to adaptive learner interfaces, users are looking for smart learning technology systems that provide a highly personalised user experience. As a result, there is a significant need to redefine their conventional development. Artificial Intelligence (AI) can solve these challenges and implement innovative digital techniques and tools in education. AI techniques can drive efficiency, personalization, and customization of the experience for different learning groups, teachers, and instructors. AI makes educational software more user-oriented, helps in the realization of complex tasks and processes of huge data, reducing their execution time and optimizing the overall functionality of the system. Individualized schedules, customized tasks, interaction with digital technologies and personal recommendations are part of the personal approach of each learner who uses AI. Besides, a personal approach helps learners feel special, increasing their engagement and raising interest in studies in such a way.

This session aims to address the research on high-quality, high-impact, original research results reporting the current state of the art of online education systems empowered with AI techniques (e.g., machine/deep learning, knowledge representation, neural networks, reinforcement learning, fuzzy logic, cognitive maps, genetic algorithms, natural language processing, social intelligence, etc.). It offers a forum for the constructive interaction and prolific exchange of ideas among scientists and practitioners on a range of research fields targeting automated intelligent support in education applications, covering different levels of the intelligent techniques and their applications in educational process with special attention but not limited to the following topics:

• Personalization and Adaptation in e-Learning Systems

• Modern Learning Paradigms and Artificial Intelligence

• Smart Web-Based Learning

• Intelligent Tutoring Systems

• Intelligent Textbook

• Application of Recommendation Techniques in E-learning Environments

• Intelligent Educational Systems in Different Domains

• Contemporary Paradigms (Virtual Reality, Augmented Reality, Eye-Tracking) as a Support in Educational Platforms

• Authoring Tools for Intelligent Tutoring Systems

• Learning with AI Systems

• Agent-Based Learning Environments

• Intelligent Techniques (e.g. Deep Learning, Fuzzy Logic, Neural Networks, Genetic Algorithms, Reinforcement Learning, Cognitive Maps, etc.) in Education

• Architectures for AI-based Educational Systems

• Learning Analytics and Educational Data Mining

• Educational Robotics

• Intelligent and Interactive Technologies in an Educational Instructional Design

Organisers

Aleksandra Klasnja-Milicevic, University of Novi Sad, Serbia. Email: akm@dmi.uns.ac.rs

Mirjana Ivanovic, University of Novi Sad, Serbia. Email: mira@dmi.uns.ac.rs

Scope and Topics

Data selection focuses on reducing the training time and, at the same time, taking advantage to do better predictions. Too much information is not handy at all since uninformative samples or features may be learnt and consequently the ability to generalize could be hindered. Addressing any problem may mean not having prior knowledge and even to become able, through data selection and even transformation measure, to learn the important data for the forthcoming prediction on unseen data. Depending on the followed methodology to conduct the process model for data mining, the data selection may be named with different names although the core is the same. Tools based on graphical user interfaces are of particular interest in the sense that may make easier the procedure to refine the raw data and eventually to get the ready data to face the mining phase. Data pre-processing deals with many tasks such as data cleansing, attribute selection, instance selection, noise reduction and detecting wrong or distorted labels. Visual data analytics is on the rise especially in multi-dimensional business applications. We encourage to submit very recent applications and if possible unprecedented. Additionally, new theoretical or empirical approaches are welcome. The topics of interest for this session include, but are not limited to:

• Data selection

• Data pre-processing

• Data cleansing

• Data engineering

• Attribute selection

• Instance selection

• Data fusion

• Data mining

• Text mining

• Speech mining

• Signal mining

• Stream mining

• Motif mining 

• Itemset mining

• Sequential pattern mining 

• Frequent pattern mining

• Infrequent pattern mining

• Rare pattern mining

Organisers

Antonio J. Tallón-Ballesteros, University of Huelva, Spain. Email: antonio.tallon.diesia@zimbra.uhu.es

Ireneusz Czarnowski, Gdynia Maritime University, Poland. Email: i.czarnowski@umg.edu.pl

Scope

Technology has been rapidly changing educational paradigms and environments. With the COVID-19 pandemic, the increased use of technologies such as videoconferencing tools supported by Learning Management Systems (LMS) soared among educational institutions, posing new challenges with respect to traditional classroom approaches. As a matter of fact, technology in education can be disruptive to the teacher-centered education and promote a new view focusing on problem solving and hands-on activities, using mobile phones and robotics, or addressing new ways of promoting students' active participation, as in flipped classrooms.  On the other hand, the internet and new intelligent technologies offer novel ways for teachers to scaffold learning activities and to provide more personalized tutoring to students. 

This workshop is intended to address new paradigms emerging using educational technology, and to present examples of how technology has been a drive for change in formal and informal learning environments.  Some examples of the new technologies include the use of Learning Management Systems, in synchronous and asynchronous modes.

Learning Objects Repositories (LORs) are another relevant technology used as tools to enhance learning. Their main features are the reusability and retrieval capacities making them not only potentially ubiquitous but also easily sharable and updatable. They are an indispensable tool with respect to the next generation of web technologies, the web 3.0, supporting web semantics. By using metadata, LOs are easily catalogued and can be easily identified and referenced on a semantic web, providing new innovative ways to support learning. Which are the best practices for using LOs in LORS? How can we motivate a learning community to produce and use LOs? Which are the best ways to integrate LOs in the planning of classroom activities?

Along with the scaffolding strategy to leverage better learning in the classroom, LORS are becoming pervasive. One example is the use of games or simulations in classes. More general approaches have been addressed using Artificial Intelligence systems, modelling learners or groups and communities and, even, using intelligent techniques to support disadvantaged schools and students. How can these disruptive technologies enhance the learning outcomes? What are the expected transformations to occur in classes using these technologies? 

Topics

Artificial Intelligence in Education, Educational Robotics, Repositories for Learning Activities, Gamification, LMS in synchronous and asynchronous classes, Learning Objects (LOs), Adaptive and Intelligent educational systems, Creative Computing, IoT Applications for Education, ubiquitous learning, linked-data and enhanced learning, Technology Enhanced Learning, Learning Analytics, Social Networks and Education, Assessment and Testing in TEL, Portals and Platforms for smart e-learning, knowledge modeling, intelligent tutoring systems (ITS), Mobile learning, Augmented and Virtual Reality in education 

Organisers

Armando B. Mendes, Universidade dos Açores, Portugal. Email: armando.b.mendes@uac.pt

José M. V. R. Cascalho, Universidade dos Açores, Portugal. Email: jose.mv.cascalho@uac.pt

Scope

This special session aims at  encompassing the latest innovations about data mining in the context of Finance. Therefore, the presentation of works tackling theoretical issues and applications, from industry or academia, on data mining is welcome. The finance may be focused on short, medium or long-term period; all of them have room in this session. The problem complexity is increasing due to the huge amount of transactions that are happening at a specific instant all over the world and is not very easy to detect the causality among all of them independently of the very distant point wherever the operations could take place.

Topics

The topics of interest for this session include, but are not limited to:

• Finance

• Financial data mining

• Economical prediction

• Stock-market analysis

• Markets on the rise

• Reputation on finance

• Economist intelligence

• Money laundering avoidance

• Early detection of tax evasion

• Technological issues affecting the economy of a company

• Big numbers in the economy of a country

• Contemporary trade

• Business intelligence

Organisers

Paulo Vasconcelos, University of Porto, Portugal. Email: pjv@fep.up.pt 

Antonio J. Tallón-Ballesteros, University of Huelva, Spain. Email: antonio.tallon.diesia@zimbra.uhu.es