European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning

Call for papers
Special sessions
Author guidelines
Infos speakers
Social events
Get a password

Login:  Password:   Password forgotten ?

Special sessions

Special sessions are organized by renowned scientists in their respective fields. Papers submitted to these sessions are reviewed according to the same rules as any other submission. Authors who submit papers to one of these sessions are invited to mention it on the author submission form; submissions to the special sessions must follow the same format, instructions and deadlines as any other submission, and must be sent according to the same procedure.

The following special sessions will be organized at ESANN2018:

Deep Learning in Bioinformatics and Medicine
Organized by Miguel Atencia (Universidad de Málaga, Spain), Davide Bacciu (Università di Pisa, Italy), Paulo J. G. Lisboa (Liverpool John Moores University, United Kingdom), Jose D. Martin, (Universitat de València, Spain), Ruxandra Stoean (University of Craiova, Romania), Alfredo Vellido (Universitat Politècnica de Catalunya, Spain)

Deep learning (DL) has been harnessing the attention of the machine learning research community over the latter years. Much of its success roots on having made available models and technologies capable of achieving ground-breaking performances in a variety of traditional fields of application of machine learning, such as machine vision and natural language processing (NLP).

Medicine, genetics, biology and chemistry are among the research fields where machine learning models find most consolidated applications. Admittedly, some of the DL flagships, like NLP and image processing have their implications in Medicine, e.g., in extracting information from the text of patients’ records or in analyzing medical imagery to find anomalous patterns. However, deep learning methodologies have only recently started to be used to address relevant bioinformatics and cheminformatics challenges. Reasons for such a slowed-down permeation can be sought in the complexity of the deep learning models which might prove difficult to use in novel application fields by non-machine learning experts. Lack of interpretability and insight into the trained models might also have been a limiting factor.

Despite such few limitations, deep learning methodologies offer far more enabling aspects and technologies for developing impacting contributions in bioinformatics research. Between the most relevant are the ability to effectively and efficiently process complex, large scale and multi-modal data, e.g. collections of biomedical images and associated patient information, DNA sequences, molecular graphs. The modular design of deep architectures together with the potential for re-using parts of previously trained models on novel tasks is another potential success enabler for bioinformatics applications.

This special session is meant to attract researchers who develop, investigate, or apply deep learning methods on biomedical and chemistry data. We aim to bring together researchers working on the topic from both the deep learning and the bioinformatics communities.

Topics include, but are not restricted to:

  • Deep learning applications and novel models for biology, chemistry, genetics, medicine and omics-data
  • Interpretability and provable properties of deep learning models
  • Learning representations from multi-modal bioinformatics data
  • Deep models for visual analytics and inspection of biomedical data
  • NLP for knowledge discovery in the medicine field
  • Deep Reinforcement Learning for the optimization of medical treatments
  • Deep learning for structured data processing in bioinformatics and chemistry
  • High performance computing for deep learning and bioinformatics
  • Software frameworks and toolkits specific for deep learning in bioinformatics and medical applications

Machine Learning and Data Analysis in Astroinformatics
Organized by Michael Biehl, Kerstin Bunte (University of Groningen, The Netherlands), Giuseppe Longo (University of Naples, Italy), Peter Tino (University of Birmingham, United Kingdom)

The ever-growing amount of data which becomes available in many domains clearly requires the development of efficient methods for data mining and analysis. These challenges occur in a variety of areas including societal issues, business and fundamental scientific research.

Astronomy continuous to be at the forefront of this development: Modern observational techniques provide enormous amounts of data, which have to be processed efficiently. The development of methods for their reliable acquisition and analysis has immediate impact on other areas including commercial applications, data security, environmental monitoring etc.

This special session is meant to attract researchers who develop, investigate or apply methods of machine learning and data analysis in the context of astronomical data.

Potential topics include, but are not limited to

  • big data mining in astronomy
  • the processing of astronomical images
  • filtering techniques for streams of astronomical data
  • outlier and novelty detection in observational data
  • classification or clustering of celestial objects
  • simulation of astrophysical models and related
  • inference problems
  • the analysis of heterogeneous data stemming from
  • various sources or technical platforms

Interaction and User Integration in Machine Learning for Information Visualisation
Organized by Bruno Dumas, Benoit Frénay (Université de Namur), John Lee (Université catholique de Louvain, Belgium)

Many methods have been developed in machine learning (ML) for information visualisation (infovis). For example, PCA, MDS, t-SNE and improvements are standard tools to reduce the dimensionality of high dimensional datasets for visualisation purposes. However, multiple other means are regularly used in the field of infovis when tackling datasets with high dimensionality. Letting the user manipulate the visualisation is one of these means, either through selection, navigation or filtering. Introducing manipulation of the visualisation also integrates the user as a core aspect of a given system. In the context of machine learning, beyond the informational and exploratory use of infovis, users' feedback can for example be highly informational to drive the dimensionality reduction process.

This special session of the ESANN conference is a followup of the special session on "Information Visualisation and Machine Learning: Techniques, Validation and Integration" at ESANN 2016. It aims to gather researchers that integrate users in the core of ML methods for infovis. New algorithms and frameworks are welcome, as well as experimental use cases that bring new insight in the integration of interaction and user integration in ML for infovis. This special session aims to provide practitioners from both communities a common forum of discussion where issues at the crossroads of machine learning and information visualisation could be discussed.

Topics of interest include, but are not limited to:

  • supervised and semi-supervised machine learning and infovis
  • unsupervised ML (clustering, dimension reduction)
  • user feedback on metaparameters
  • new visual paradigms for machine learning
  • interaction techniques for infovis with/of machine learning
  • user and device adaptivity for visual analytics
  • warm restart and dedicated optimization techniques
  • scalability
  • applications in industry, agriculture, medicine, biology, etc.

Emerging trends in machine learning: beyond conventional methods and data
Organized by Luca Oneto (University of Genoa), Nicolò Navarin (University of Padua), Michele Donini (Istituto Italiano di Tecnologia), Davide Anguita (University of Genoa, Italy)

Recently, new promising theoretical results, techniques, and methodologies have attracted the attention of many researchers and have allowed to broaden the range of applications in which machine learning can be effectively applied in order to extract useful and actionable information from the huge amount of heterogeneous data produced everyday by an increasingly digital world.

Examples of these methods and problems are:

  • Learning under privacy and anonymity constraints
  • Learning from structured, semi-structured, multi-modal (heterogeneous) data
  • Constructive machine learning, e.g. generative models and structured output learning
  • Reliable machine learning
  • Learning to learn, e.g. lifelong learning and learning the loss
  • Mixing deep and structured learning, e.g. mixture of wide and deep models
  • Semantics-enabled recommender systems
  • Reproducibility and interpretability in machine learning
  • Human in the loop
  • Adversarial learning
The focus of this special session is to attract both solid contributions or preliminary results which show the potentiality and the limitations of new ideas, refinements, or contaminations between the different fields of machine learning and other fields of research in solving real world problems. Both theoretical and practical results are welcome to our special session.

Shallow and Deep models for transfer learning and domain adaptation
Organized by Siamak Mehrkanoon, Matthew Blaschko, Johan Suykens (KU Leuven, Belgium)

Manual labeling of sufficient training data for diverse application domains is a costly, laborious task and often prohibitive. Therefore, designing models that can leverage rich labeled data in one domain and be applicable to a different but related domain is highly desirable. In particular, domain adaptation or transfer learning algorithms seek to generalize a model trained in a source domain (training data) to a new target domain (test data). The most common underlying assumption of many machine learning models is that both training and test data exhibit the same distribution or the same feature domains. However, in many real life problems, there is a distributional, feature space and/or dimension mismatch between the two domains or the statistical properties of the data evolve in time.

Transferring and incorporating different sources of information such as learned feature extractors, knowledge of labeled and unlabeled instances, learned parameters among others from different domains into a unified model that can leverage all the available prior knowledge in order to achieve human level accuracy on a given new task is of great importance. In this context, depending on the availability of the labeled and unlabeled training data from (i) source domains, (ii) source and target domains, different scenarios related to supervised as well as semi-supervised domain adaptation can for instance be considered. In addition different modeling strategies ranging from shallow to deep models is of interest.

Therefore, the main objective of the session is to discuss the recent rise of new research questions and learning strategies for the domain adaptation and transfer learning problems using both shallow and deep models. The goal is to promote a fruitful exchange of ideas and methods between different communities, leading to a global advancement of the field.

The topics of interest include but not limited to:

  • Deep and shallow models
  • Neural Networks
  • Kernel based models
  • Transfer learning and domain adaptation
  • Feature learning / representation learning
  • Domain invariant features
  • Supervised / Semi-supervised Learning
  • Fine-tuning / Feature extractor / Amount of labelling
  • Scalability
  • Regularization

Randomized Neural Networks
Organized by Claudio Gallicchio, Alessio Micheli (University of Pisa, Italy), Peter Tino (University of Birmingham, United Kingdom)

The use of randomization in the design of Neural Networks (NNs) has become increasingly popular, mainly due to the ease of implementation, extreme efficiency of the training algorithms and the possibility of analyzing the NNs properties that are independent from learning. Randomization can enter NN design in various disguises, for example in the model construction and training (e.g. random setting of a subset of weights), or in its functionality and regularization algorithms (e.g. inclusion of random noise in activation layers, drop-out techniques, etc.). Under a broader perspective, the analysis of randomized models naturally extends to a general Machine Learning (ML) context (e.g. random projections). Moreover, Learning in Structured Domains and Deep Learning represent ML research areas for which this type of analysis is highly beneficial.

This session calls for contributions targeting novel theoretical and/or empirical studies on randomization in NNs, and it is proposed as an opportunity for discussing the advantages and limitations/shortcomings of the approach under an open and critical perspective.

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

  • Neural Networks with random weights
  • Randomized Machine Learning algorithms
  • Reservoir Computing and Echo State Networks
  • Extreme Learning Machines and Random Vector Functional-Link Networks
  • Random Projections and Neural Networks
  • Randomized regularization techniques
  • Bias of randomization in the design of Neural Networks
  • Theoretical analysis: advantages and shortcomings (range of applicability, stability, efficiency, etc.)
  • Deep Randomized Neural Networks
  • Randomized approaches for Learning in Structured Domains (sequences, trees, graphs)
  • Efficient implementations of Randomized Neural Networks
  • Applications and comparisons

Impact of Biases in Big Data
Organized by Patrick Glauner (University of Luxembourg), Petko Valtchev (University of Quebec at Montreal, Canada), Radu State (University of Luxembourg)

For about the last decade, the Big Data paradigm that has dominated research in machine learning can be summarized as follows: "It’s not who has the best algorithm that wins. It’s who has the most data." However, most data sets are biased and the corresponding biases are often ignored in research. This in turn makes the learned models unreliable. Concretely, the most frequently appearing biases in data sets are class imbalance and sample selection bias.

Class imbalance refers to a data set having a substantially different amount of examples per class. Models trained on such data sets often tend to predict the majority class, e.g. when using inappropriate metrics such as the accuracy. Sample selection bias refers to the problem of training data and production data having different distributions. In many real-world applications, this is a common issue because we do not have complete control over the data gathering process. Models trained on such training data poorly generalize to the production data. For both class imbalance and selection bias, having more representative data will help rather than just having a lot of more data.

This special session will be an opportunity for both researchers and practitioners, to present and discuss their latest works on the impact of biases in data sets on models and how these can be reduced. Topics include, but are not limited, to:

  • Quantifying biases
  • Class imbalance
  • Sample selection bias and covariate shift
  • Global and local learners
  • Biases in Deep Learning
  • Evaluation metrics for biased data sets
  • Reweighting and subsampling methods
  • Biases in anomaly detection
  • Biases in spatial data and time series
  • Sociological impact of biased machine learning models, e.g. biases in the criminal justice system

For any information: Michel Verleysen -