December 13, 2016

Abstracts

Artificial Intelligence in Nanophotonics Workshop

Dr. Nikolay Zheludev

Dr. Cesare Soci

University of Southampton and NTU, Singapore

Abstract

The workshop at the 15th International Work-Conference on Artificial Neural Networks, IWANN2019, June 12-14, 2019 in Gran Canaria, (Canary Islands, Spain) will feature original recent results and topical reviews on:

  1. The proliferation of AI approaches into the design of nanodevices, optimization of functionalities of artificial photonic nanostructures and materials
  2. Neuromorphic photonic systems for artificial cognition and analog computing

SS01: Artificial Neural Network for biomedical image processing

Dr. Yu-Dong Zhang

University of Leicester, UK

Abstract

With advancement in biomedical imaging, the amount of data generated by multimodality image techniques (e.g. ranging from Computed Tomography (CT), Magnetic Resonance Imaging (MR), Ultrasound, Single Photon Emission Computed Tomography (SPECT), and Positron Emission Tomography (PET), to Magnetic Particle Imaging, EE/MEG, Optical Microscopy and Tomography, Photo acoustic Tomography, Electron Tomography, and Atomic Force Microscopy, etc.) has grown exponentially and the nature of such data is increasingly become more complex. This poses a great challenge on how to develop new advanced imaging methods and computational models for efficient data processing, analysis and modelling in clinical applications and in understanding the underlying biological process.

The purpose of this special issue aims to provide a diverse, but complementary, set of contributions to demonstrate new developments and applications of advanced imaging analysis in the multimodal biomedical imaging area. The ultimate goal is to promote research and development of advanced imaging analysis for multimodal biomedical images by publishing high-quality research articles and reviews in this rapidly growing interdisciplinary field.

  • New algorithms, models and applications of advanced imaging methods
  • Multimodal imaging techniques: data acquisition, reconstruction; 2D, 3D, 4D imaging, etc.)
  • Translational multimodality imaging and biomedical applications (e.g., detection, diagnostic analysis, quantitative measurements, image guidance of ultrasonography)
  • Variational and combinatorial optimizations for biomedical imaging and image analysis
  • Advanced Biomedical image analysis (image processing, Statistical and probabilistic methods for biomedical imaging and image analysis, Machine learning in biomedical imaging and image analysis)
  • Deep learning methods (convolutional neural network, auto encoder, deep belief network, etc.)
  • Visualization

SS02: Deep learning models in healthcare and biomedicine

Dr. Leonardo Franco

Universidad de Málaga, Spain

Dr. Ruxandra Stoean

University of Craiova, Romania

Dr. Francisco Veredas

Universidad de Málaga, Spain

Abstract

Medicine seems like the perfect candidate for deep learning exploration. As such, there is a continuous interest in tailoring deep networks to discover the complex interactions and models within the field. This session welcomes recent contributions in all areas connected to the design and application of deep learning techniques for healthcare tasks and biomedical studies, including, but not limited to:

  • Deep learning models in healthcare:
    • Convolutional neural networks for medical image processing
    • Convolutional and LSTM networks for sequence processing of medical records
    • Deep learning tools for diagnosis and grading of disease
    • Interpretation of deep learning models in healthcare tasks
    • Future challenges in deep learning healthcare applications
  • Deep learning models in biomedicine:
    • Deep learning models in bioinformatics
    • Auto-encoder models for genomics, proteomics and multi-omics studies
    • Convolutional neural network approaches for biomedical data
    • Recurrent neural network approaches in biomedicine
    • Data augmentation techniques for genomics and proteomics applications
    • Transfer learning strategies in bioinformatics
    • Deep learning models in precision medicine

SS03: Deep learning beyond convolution

Dr. Miguel Atencia

Universidad de Málaga, Spain

Abstract

The description of deep learning architectures is often simplified to include only supervised learning on feedforward networks, i.e. multilayer perceptrons alternating backpropagation and convolutional layers. However, the successful performance of deep learning architectures can arguably have been boosted by supplementary modules providing preprocessing, reshaping, normalization, and dimensionality reduction. In this session, we aim at exploring such alternative architectures, either on their own, or as layers of a more complex network. These include, but are not limited to:

  • Recurrent models, such as
    • Long Short Term Memory Networks
    • Echo State Networks
  • Dimensionality reduction techniques, in a wide sense, e.g.:
    • Autoencoders
    • Generative Adversarial Networks
    • Variational autoencoders
    • Nonnegative Matrix and Tensor Factorization
    • Feature selection and clustering algorithms

Contributions including theoretical analysis of such hybrid Deep Learning models performance, e.g. in the context of manifold learning, as well as successful applications to classification and optimization tasks, are particularly welcome.

SS04: Machine learning in Vision and Robotics

Dr. José García-Rodríguez

University of Alicante, Spain

Dr. Enrique Domínguez

University of Malaga, Spain

Dr. Ramón Moreno

Lortek, Spain

Abstract

Over the last decades there has been an increasing interest in using machine learning and in the last few years, deep learning methods, combined with other vision techniques to create autonomous systems that solve vision problems in different fields. This special session is designed to serve researchers and developers to publish original, innovative and state-of-the art algorithms and architectures for real time applications in the areas of computer vision, image processing, biometrics, virtual and augmented reality, neural networks, intelligent interfaces and biomimetic object-vision recognition.

This special session provides a platform for academics, developers, and industry-related researchers belonging to the vast communities of Neural Networks, Computational Intelligence, Machine Learning, Deep Learning, Biometrics, Vision systems, and Robotics , to discuss, share experience and explore traditional and new areas of the computer vision, machine and deep learning combined to solve a range of problems. The objective of the workshop is to integrate the growing international community of researchers working on the application of Machine Learning and Deep Learning Methods in Vision and Robotics to a fruitful discussion on the evolution and the benefits of this technology to the society.

The methods and tools applied to vision and robotics include, but are not limited to, the following:

  • Computational Intelligence methods:
    • Machine Learning methods
    • Self-adaptation and self-organisation
    • Robust computer vision algorithms (operation under variable conditions, object tracking, behaviour analysis and learning, scene segmentation)
    • Extraction of Biometric Features (fingerprint, iris, face, voice, palm, gait)
    • Convolutional Neural Networks CNN
    • Recurrent Neural Networks
    • Deep Reinforcement learning

The fields of application can be identified, but are not limited to, the following:

  • Video and Image processing
  • 3D Scene reconstruction and Volume visualization
  • 3D Tracking in Virtual Reality Environments
  • Intelligent Interfaces (User-friendly Man Machine Interface)
  • Multi-camera and RGB-D camera systems
  • Multi-modal Human Pose Recovery and Behavior Analysis
  • Gesture and posture analysis and recognition
  • Biometrics and Surveillance systems
  • Autonomous and Social Robots
  • Industry 4.0
  • IOT and Ciberphysical Systems

SS05: Data-driven Intelligent Transportation Systems

Dr. Ignacio J. Turias Domínguez

Universidad de Cádiz, Spain

Dr. David Elizondo

De Montfort University, Leicester, UK

Dr. Francisco Ortega Zamorano

Universidad de Málaga, Spain

Abstract

Intelligent Transportation Systems (ITS) use the Internet of Things (IoT) and advanced data communication technologies to build an integrated system of people, vehicles and roads. Accurate and real-time traffic related data is required to improve the performance of transportation systems, to monitorize and/or forecast the pollution rate, or to detect traffic congestion or disruption, among many others. In the recent years, the fast growth in technology has allowed the collection and availability of heterogeneous traffic related data, which enables the application of cutting-edge data analytics tools to overcome the latest challenges in this field. This session welcomes recent contributions in all áreas connected to data-driven techniques in ITS which address transport related issues including, but not limited to:

  • Intelligent Transportation Systems and air pollution:
    • Neural networks for air quality forecast.
    • Prescriptive data analysis for local authorities’ assessment.
    • Vehicular air pollution in local urban areas.
    • Relationship among multimodal transportation and air pollution.
    • Freight distribution in urban contexts.
    • Maritime transportation and air pollution
  • Computational intelligence in transportation systems:
    • Traffic monitoring and surveillance using convolutional neural networks.
    • Neural networks for short-term traffic or flow forecast.
    • Disruption and/or disaster management systems.
    • Driver assistance systems, autonomous cars.
    • Route guidance and optimization through local and real-time vehicle
      information.
    • Intelligent containerization.
    • ITS for international transport (e.g., cross-border and transit transport
      facilitation) and intermodal transport.
  • Sensor and network technology for Intelligent Transportation Systems:
    • IoT infrastructure to enhance traffic related data acquisition.
    • Computer vision.
    • Cybersecurity.
    • Wireless sensor networks and communications in vehicles.
    • Smart sensors: edge and/or fog computing.
    • Data fusion methods for heterogeneous sensors

SS06: Software Testing and Intelligent Systems

Dr. Juan Boubeta

Universidad de Cádiz, Spain

Dr. Pablo C. Cañizares

Universidad Complutense de Madrid, Spain

Dr. Gregorio Díaz

Universidad de Castilla-La Mancha, Spain

Abstract

Current software systems are increasingly complex and, therefore, it is more difficult and costly to ensure that they do what they are supposed to do. Software testing plays a key role to increase the confidence on the correctness of systems. Despite the huge amount of resources devoted to testing (up to 60% of the budget), testing is still mainly a manual and prone to errors process. Therefore, there is a need to improve testing so that costs can be cut, by automating most of the tasks, and the amount of detected errors can be increased, by using better techniques.

Intelligent systems are ubiquitous in our daily routine: smartphones, navigation systems, smartwatches, etc. In addition to be the basis of (more or less) sophisticated gadgets, these systems are fundamental in areas such as healthcare diagnostics and medical devices, traffic estimation, weather forecast, and many others. They are a clear case of complex systems. Therefore, these systems are difficult to design, implement, and test. In the case of testing, classical techniques cannot be used because intelligent systems have some peculiarities. On the one hand, they are usually governed by non-deterministic algorithms using advance AI techniques where classical testing will struggle. On the other hand, they have to analyze huge amounts of data in real-time, so that it is extremely important that the solutions scale. Therefore, it is important that testing methodologies adapt to deal with these challenging systems, so that the number of errors can be reduced, avoiding recalls derived from wrong implementations.

During the last years, we are contemplating the emergence of new testing techniques based on the application of AI techniques such as evolutionary computation, artificial life, neural computation and swarm intelligence. Therefore, there is a feedback process between the fields: the reliability of intelligent systems is improved thanks to good software testing methodologies and software testing is improved thanks to knowledge obtained from the techniques used to develop intelligent systems. The main aim of this special session is to contribute to the progress in the improvement and appropriate use of software testing and intelligent systems. We are interested in the adaption of existing testing approaches, as well as new ones, to test intelligent systems. In addition, we look forward to novel testing techniques based on computational intelligence paradigms. We are sure that the collaboration of researchers from different areas will result in benefits that can be applied in some of the research lines that are under the umbrella of the IWANN conference.

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

  • Heuristic techniques in software testing.
  • Formal approaches in intelligent systems.
  • Risk analysis of intelligent systems.
  • Swarm intelligence in software testing.
  • Monitoring of intelligent systems.
  • Case studies and applications

SS07: Deep Learning and Natural Language Processing

Dr. Leonor Becerra-Bonache

University of Saint-Etienne, France

Dr. M. Dolores Jiménez-López

Universitat Rovira i Virgili, Spain

Dr. Benoit Favre

Aix-Marseille Université, France

Abstract

This special session focuses on the common space delimited by two areas: natural language processing and deep learning. The main goal of the special session is to promote interdisciplinarity among people working in such disciplines, boosting the interchange of knowledge and viewpoints between specialists. This interdisciplinary research can provide new models that may improve AI technologies.

The large amount of text generated by humans every year has increased the need to develop computers interfaces with the ability to communicate with humans in natural languages. Natural Language Processing (NLP) is a subfield of Artificial Intelligence that focuses on enabling computers to process and understand natural languages. It has become one of the most important technologies in the information age.

For a long time, NLP techniques were dominated by linear modeling approaches to supervised learning (e.g., linear SVM or logistic regression). However, NLP is being revolutionized by deep learning approaches, with which have been achieved superior results across many different NLP tasks as compared to traditional machine learning approaches. This special session aims to bring together experts in deep learning and natural language processing that present their new results in applying deep learning methods for solving different NLP tasks, discuss about the impact of deep learning on the field of NLP and debate on the limitations of deep learning in NLP. We will also encourage works that present new developments in applying NLP for solving problems related to Deep Learning.

We think that our proposal can be very relevant to the main conference, since it focuses on one of the IWANN 2019 trending topics (deep learning) and it may complement the regular program with a topic such as Natural Language Processing (not usually represented in the conference) that can be of interest to the participating community since it may be one of the natural applications of neural computing. One of the important values of the special sesión is its multi-disciplinary character. In fact, the main objective of this session is to promote the exchange of ideas, challenges and perspectives between specialists that working on NLP or in deep learning have interest in using methods from other disciplines that can provide new ideas, new tools and new formalisms to approach their problems and that can help in the improvement of their theories and models. Because it affords a cutting-edge topic and adopts a clearly multidisciplinary pespective, we are convinced that IWANN 2019 can clearly benefit from the inclusion of a special session of this type attracting a new/different kind of participants/researchers.


SS08: Random-Weights Neural Networks

Dr. Claudio Gallicchio

University of Pisa, Italy

Abstract

Random-weights Neural Networks identify a class of artificial neural models that employ a form of randomization in both their architectural and training design. Typically, connections to the hidden layer(s) are left untrained after initialization, and only the output weights need to be adjusted through learning (typically, by means of non-iterative methods). Extreme efficiency of training algorithms, along with the ease of implementation, made the randomized approach to Neural Networks design an incredibly widespread and popular methodology among both researchers and practitioners. Besides, from a theoretical perspective, randomization enables an effective study of the inherent properties for various kinds of Neural Networks architectures, even in the absence of (or prior to) training of internal weights connections. In literature, the approach has been instantiated in several forms, both in the case of feed-forward models (e.g., Random Vector Functional Link, Extreme Learning Machine, No-prop and Stochastic Configuration Networks), and for recurrent architectures (e.g., Echo State Networks, Liquid State Machines). Moreover, the rise of the Deep Learning era in Machine Learning research has recently given a further impulse to the study of hierarchically organized neural architectures with multiple random-weights components. In this concern, the potentialities of combining the advantages of deep architectures and the efficiency of randomized Neural Networks approaches remain still largely unexplored.

This session calls for contributions in the area of random weights Neural Networks from all perspectives, from seminal works on breakthrough ideas to applications of consolidated learning methodologies. Topics of interest for this session include, but are not limited to, the following:

  • Neural Networks with random weights
  • Randomized algorithms for Neural Networks
  • Non-iterative methods for learning
  • Random Vector Functional Link, Extreme Learning Machines, No-prop, and Stochastic Configuration Networks
  • Reservoir Computing, Echo State Networks, and Liquid State Machines
  • Deep Neural Networks with Random Weights (e.g. Deep Extreme Learning Machines and Deep Echo State Networks)
  • Theoretical analysis on advantages and downsides of randomized Neural Networks
  • Comparisons with fully trained Neural Networks
  • Real-world Applications

SS09: New and future tendencies in Brain-Computer Interface systems

Dr.Ricardo Ron Angevín

University of Malaga, Spain

Dr. Ivan Volosyak

Rhine-Waal University of Applied Sciences, Germany

Abstract

A Brain-computer Interface (BCI) is based on the analysis of brain activity to provide a non-muscular channel for sending messages and commands to the external devices. A BCI emerged few decades ago as a new communication channel allowing subjects with severe neuromuscular disorders, who may be completely paralyzed or locked-in, to communicate and to interact with the outer world. However, modern BCI applications have been also used in totally different areas (e. g. entertainment). In the future, a further qualitative improvement in performance is expected, in terms of information transfer rate and reliability with potential uses in emerging areas of interest.

The main goal of this special session is to show the last research advances in neurotechnologies, BCIs and applications, as well as innovative signal processing algorithms, training techniques or paradigms that make them possible.

The organizers have successfully organized five special sessions about BCIs technologies in last IWANN conferences: IWANN09 (Salamanca), IWANN11 (Málaga), IWANN13 (Tenerife), IWANN15 (Palma de Mallorca) and IWANN17 (Cádiz). In this sense, the BCI special session is a traditional special session in IWANN conference, allowing different researchers in the BCI field to meet every two years.


SS10: Human Activity Recognition

Dr.-Ing. habil. Matthias Pätzold

University of Agder, Norway

Abstract

Human activity recognition is an important research area in computer vision for various contexts including healthcare, security surveillance, and human computer interaction. Given the breath of potential applications, it is no surprise that the research demand in the area of human activity recognition has grown rapidly in recent years. The primary objective of this special session is to discuss the key features and challenges of activity recognition. Another objective is to bring together people working on theoretical and practical aspects of human activity recognition. The latest breakthroughs in this area will be reported and future research directions will be discussed.

Potential authors are invited to submit papers describing original contributions to human activity recognition. Thereby, theoretical, computational, as well as experimental studies related to activity recognition are welcome. The scope of this special session includes, but is not limited to, the following topics:

  • Deep learning for sensor-based human activity recognition
  • Machine learning for sensor-based human activity recognition
  • Multi-user activity recognition
  • Wearable sensor-based activity recognition
  • Radio frequency (RF-)based activity recognition
  • Localization of human activities
  • Combination of wearable and non-wearable sensors
  • Feature-selection and combination of features
  • Emotion and intention recognition
  • Modelling and simulation of human activities.

SS11: Computational Intelligence Methods for Time Series

Dr. Héctor Pomares

University of Granada, Spain

Abstract

Within the field of science and engineering, it is very common to have data arranged in the form of time series data which must be subsequently analysed, modelled and classified with the eventual goal of predicting future values. The literature shows that all these tasks related to time series can be undertaken using computational intelligence methods. In fact, new and further computational intelligence approaches, their efficiency and their comparison to statistical methods and other fact-checked computational intelligence methods, is a significant topic in academic and professional projects and works. Therefore, this special session aims at showing to our research community high quality and state of the art computational intelligence (and statistical) related works, applied to time series data and their tasks: analysis, forecasting, classification, and clustering. Furthermore, the experts can, from the starting point that the works shown provide, discuss different solutions and research issues for these topics.

The topics of interest include but are not limited to:

Computational Intelligence (CI) techniques applied to

  • time series analysis
  • time series modeling
  • time series forecasting
  • time series classification
  • time series clustering
  • statistical and CI techniques for time series, comparative evaluation and/or novel propositions

SS12: Advanced Methods for Personalized/Precision Medicine

Dr. Luis Javier Herrera

Dr. Fernando Rojas

University of Granada, Spain

Abstract

It is becoming astonishing the way advances in ICT are changing not only our daily life, but the way all the rest of sciences and research areas are growing and evolving, Bioinformatics, Biotechnology and Biomedicine among them (or in general Personalized/Precision Medicine). These increasing changes have reached a new milestone with the advent of Deep Learning (DL). This new paradigm has set up a countless number of possibilities and solutions for new and already well-known applications. Its potential is well reflected in the overcoming of human capabilities in very complex tasks such as image and video recognition, and audio processing and synthesis, among others. However, despite the readiness of the “big” human data available via these technologies, we are still at the infancy when it comes to practically exploiting all the more and more massive available information for relevant health and wellbeing applications. DL represents just one in an outstanding set of Machine Learning (ML) methodologies, which are being supported by the fast evolution of High Performance Computing (HPC) systems specifically dedicated to Big Data analysis. Powerful GPU-based architectures and Hadoop systems, together with virtualization and the use of cloud resources, represent the main computational sources allowing quick evolution and success of DL and other complex latest-generation ML algorithms trained from massive amounts of data from different and heterogeneous sources.

Topics include advances in computational intelligence methods for:

  • Genomics/proteomics for cancer and other diseases characterization
  • Improved diagnosis from images and other data
  • Personalized/precision diagnosis
  • Disease risk assessment
  • e-health advances

SS13: Exploring document information to improve neural summarization models

Dr. Luigi Di Caro

University of Torino

Abstract

Neural networks became widely used in Summarization since they can automatically generate summaries.

However, they are not able to use global information features coming from the whole document collection to improve the quality of the summary. In this session, we encourage authors to submit neural network models that deeply explore this aspect.


SS15: Machine learning in weather observation and forecasting

Dr. Juan Luis Navarro-Mesa

Dr. Antonio Ravelo-García

Dr. Carmen Paz Suárez Araujo

Universidad de Las Palmas de Gran Canaria, Spain

Abstract

Weather observations are performed using wide range of observation equipment: weather stations, wind profilers, balloons, Doppler radar, and satellites. These equipments take information about temperature, humidity, precipitation, air pressure, wind speed, wind direction, etc. These are key observations of the atmosphere that help forecasters to make predictions. Weather observations are processed and used to create forecasts. Both observation and forecasting are used in a variety of applications such as agriculture, fishing, tourism, transportation, etc. Disaster risk management is also an important field, to which, in recent times, authorities, economical agents, and people in general, are paying much attention.

Physical models of the atmosphere are traditionally used to perform weather forecasting, but the limitations of the models open new challenges. Machine learning techniques have opened a wide range of possibilities to overcome the deficiencies of traditional techniques, providing new insights to the extraction of information from the observations, classification of atmospheric phenomena and forecasting.

This session welcomes recent contributions in modeling, design and application of machine learning techniques, especially neural networks in general and deep learning in particular, in all aspects of weather information extraction and forecasting, including, but not limited to:

  • Time series analysis, modeling, forecasting, classification and clustering.
  • Feature-selection and combination of features.
  • Multi-sensor data fusión.
  • Support vector machines and kernel methods
  • Neural computing and deep learning approaches
  • New neural architectures and neural computing models: Theory and applications.