|European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning|
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 ESANN2019:
60 Years of Weightless Neural Systems
Mimicking biological neurons by focusing on the decoding performed by the dendritic trees is an attractive alternative to the integrate-and-fire McCullogh-Pitts neuron stylisation. RAM-based, or Boolean neurons, and weightless neural systems have been studied and applied in a broad spectrum of situations, resulting in theoretical findings and the development of exciting applications to an ample set of domains, ranging from natural language processing to game playing, including memory transfer mechanisms, biomedical applications, computational vision, hardware security, and quantum learning.
The year of 2019 marks the 60-years anniversary of the seminal paper on n-tuple classifiers by Bledsoe and Browning, as well as the 35-years of the WiSARD model, and the tenth anniversary of the first special session on weightless neural systems at ESANN. This session invites original contributions on theoretical and practical aspects of weightless neural systems at all levels of abstraction, as well as their relationship to themes of current interest such as: deep learning, convolutional neural models, adversarial learning, etc.
Statistical physics of learning and inference
This special session is meant to attract researchers who exploit analogies and concepts from statistical physics in the context of machine learning, inference, optimization, and related fields.
The exchange of ideas between statistical physics and computer science has been very fruitful and is currently gaining momentum again as a consequence of the revived interest in neural networks, machine learning and inference in general.
Statistical physics methods complement other approaches to the theoretical understanding of machine learning processes and inference in stochastic modeling. They facilitate, for instance, the study of dynamical and equilibrium properties of randomized training processes in model situations. At the same time, the approach inspires novel and efficient algorithms and facilitates interdisciplinary applications in a variety of scientific and technical disciplines.
The tools and concepts applied in this context include information theory, the mathematical analysis of stochastic differential equations, methods borrowed from the statistical mechanics of disorder, mean field theory, variational calculus, renormalization group and other methods.
Potential topics include, but are not limited to
Streaming data analysis, concept drift and analysis of dynamic data sets
Today many real life data are given in the form of streaming data. Prominent examples can be found in the context of IoT, in form of twitter feeds, click stream data, trading data and many other.
Learning from this huge, heterogeneous and growing amount of data requires flexible learning models that can adapt over time and are capable to deal with potentially non-i.i.d., non-stationary input data. Additionally the underlying algorithms aim on processing of high-velocity and multi-channel data and have also to deal with a variety of phenomena like concept drift and novelty detection.
This special session welcomes novel research about learning from data streams addressing common problem in the field of streaming data analysis.
Computational intelligence methods have the potential to be used for efficient data streams processing but novel methods and mathematical and algorithmic approaches are needed.
We encourage submission of papers on novel methods for streaming data processing and streaming data analysis by means of computational intelligence and machine learning approaches, including but not limited to:
Many machine learning methods are considered black box models, which makes the interpretation of decisions made by them challenging. Furthermore, these black box methods may make unforeseeable decisions for rare or unexpected input values in real-world problems. It is for these reasons that a number of domains, such as critical systems, are reluctant to move to machine learning-based decision making methods. We therefore need to bridge the gap between expert systems and machine learning methods in order to build reliable machine learning methods.
This special session will be an opportunity for both researchers and practitioners, to present and discuss their latest works on reproducible, understandable and interpretable machine learning models. Topics include, but are not limited, to:
Embeddings and Representation Learning for Structured Data
Learning models of structured data, such as sequences, trees, and graphs, has become a rich and promising research objective in many fields of machine learning, such as (deep) neural networks, probabilistic models, kernels, metric learning, and dimensionality reduction. All these seemingly disparate approaches are connected by their construction of vectorial representations and embeddings of structured data, be it implicit or explicit, fixed or learned, deterministic or stochastic. Such embeddings can not only be utilized for classification or regression, but for generation of structured data, visualization, and interpretation.
This session calls for contributions which provide novel methods to construct embeddings of structured data, new methods to utilize existing embeddings, and theoretic research regarding the properties of such embeddings. More specifically, topics of interest include, but are not limited to, the following:
Parallel and Distributed Machine Learning: Theory and Applications
The spread of Internet and the technological advances have resulted in huge volumes of data, very valuable for different agents in the industrial world that are interested in analyzing them for different purposes. Machine Learning (ML) algorithms play a key role in this context, being able to learn from and make predictions on data. Their increasing complexity, since they have to deal now with millions of parameters, as well as their computational cost lead to new research opportunities and technical challenges.
This continuous increase of data involved in ML analyses leads to a growing interest in the design and implementation of parallel and distributed ML algorithms. The efficient exploitation of the vast aggregate main memory and processing power of High Performance Computing (HPC) resources such as multicore CPUs, hardware accelerators (GPUs, Intel Xeon Phi coprocessors, FPGAs, etc.), clusters or cloud-based systems can significantly accelerate many ML algorithms. However, the development of efficient parallel algorithms is not trivial, as we must pay much attention to the data organization and decomposition strategy in order to balance the workload among resources while minimizing data dependencies as well as synchronization and communication overhead.
We invite papers on both practical and theoretical issues about incorporating parallel and distributed approaches into ML problems, as well as review papers with the state-of-art techniques and the open challenges encountered in this field. In particular, topics of interest include, but are not limited to:
Societal Issues in Machine Learning: When Learning from Data is Not Enough
It has been argued that Artificial Intelligence (AI) is experiencing a fast process of commodification. This characterization is on the interest of big IT companies, but it correctly reflects the current industrialization of AI. This phenomenon means that AI systems and products are reaching the society at large and, therefore, that societal issues related to the use of AI and Machine Learning (ML) cannot be ignored any longer.
Designing ML models from this human-centered perspective means incorporating human-relevant requirements such as safety, fairness, privacy, and interpretability, but also considering broad societal issues such as ethics and legislation. These are essential aspects to foster the acceptance of ML-based technologies, as well as to be able to comply with an evolving legislation concerning the impact of digital technologies on ethically and privacy sensitive matters.
This special session aims to put forward the state-of-the-art on these increasingly relevant topics among ML theoretician and practitioners. For this purpose, we welcome both solid contributions and preliminary relevant results showing the potential, the limitations and the challenges of new ideas, refinements, or contaminations between the different fields of research, ML, and related approaches in facing real-world problems involving societal issues.
We welcome works on ML theory, applications to topics listed below as well as other topics of social relevance. Studies stemming from major research initiatives and projects focusing on the session topics are particularly welcome.
Topics of interest
|For any information: Michel Verleysen - email@example.com|