- Low-complexity, parallel and distributed algorithms to detect and classify objects in large infrared, hyper-spectral and 3D sensor images.
- Topological analysis methods for big data.
A visual analytics approach of big data.
- Distributed control methods under communication constraints with applications to sensor networks in a smart industry setting.
- Cyber security with applications to both smart energy systems and smart industry.
- Home robotics.
Tailor-made model reduction methods for integrated energy systems.
- Modeling and control of hysteretic deformable mirrors for high-contrast imaging systems.
- Scalable algorithms to process massive datasets from radio astronomy.
- Methods for automated and robust analysis of astronomical data.
- Johann Bernoulli Institute for Mathematics and Computer Science (JBI)
- Artificial Intelligence and Cognitive Engineering (ALICE)
- Engineering and Technology Institute Groningen (ENTEG)
- Kapteyn Astronomical Institute (Kapteyn)
- Zernike Institute for Advanced Materials (ZIAM)
- Netherlands Institute for Space Research (SRON)
- Netherlands Institute for Radio Astronomy (ASTRON)
ADAPTIVE MODELS & BIG DATA
Project: Low-complexity, parallel and distributed algorithms to detect and classify objects in large infrared, hyper-spectral and 3D sensor images
This project (Bernoulli, Kapteyn) will develop low-complexity, parallel and distributed algorithms that robustly detect and classify objects in large infrared, hyper-spectral and 3D sensor images. The methodology will be based on computer vision and machine learning techniques, such as trainable morphological image processing filters which can be efficiently implemented using sequential and parallel implementations based on the max-tree or related alpha-tree data structure. We intend to explore two ways of combining these morphological scale-space data structures with machine learning, i.e. 1) using machine learning to classify nodes in the trees based on so-called vector-attribute filtering, and, 2) feeding key-point features detected by analysis of these scale spaces to deep convolutional networks, after automatic rescaling and rotation to a standard scale and orientation. This would allow scale-invariant analysis of huge images using deep learning, without excessive compute power requirements. The methods will be applied to several use cases: 1) the detection of buildings at many different types of resolution from remote-sensing data, as needed in the Global Human Settlement Layer project which aims to improve understanding of urbanisation, to assist urban planning, and 2) to support disaster relief, 3) for detection and analysis of objects in large astronomical surveys. This subproject will profit from existing collaborations of JBI with the Joint Research Centre (JRC) in Ispra, Italy. Potentially, a fourth use case on large electron micrographs obtained from the UMCG could be included.
Keywords: Scale spaces, connected filters, machine learning, remote sensing, astronomical surveys
- Prof. dr. Nicolai Petkov (JBI)
- Prof. dr. Scott Trager (Kapteyn)
- Dr. Michael Wilkinson (JBI)
- External advisor: Dr. Gijs Verdoes Kleijn (Kapteyn)
PhD student: Jiwoo You (Korea)
Project: Topological Analysis Methods for Big Data
This project is concerned with topological analysis methods for big data, and the insights this provides into the emergence of complexity in dynamical systems. The PhD student who will select this project will develop and use topological data analysis methods for the analysis of big data originating in cosmological and astronomical studies. It will be based on recent developments that have shown that state-of-the-art topological data analysis (TDA) methods, such as persistent homology, uncover new understanding and insights into the multiscale topological description of the Megaparsec weblike cosmic matter distribution. Betti numbers and topological persistence turn out to represent powerful means of describing the rich connectivity structure of the cosmic web and of its multiscale arrangement of matter and galaxies. This has shown that topological data analysis methods provide provide new means of understanding the shape of data and uncovering hidden patterns and relations. The PhD student will focus on using TDA to detect structure in data coming from large scale cosmological simulations and large cosmological redshift surveys, and use these to look into the homological properties of the observed Cosmic Web and study how these properties change in time. Intimately related to the singularity structure of the mass distribution represented in the data, the project should also enable the study of the connection with the dynamical evolution of the probed system. By means of sophisticated new reconstruction methods producing maps of the evolving mass distribution in the local Universe, the statistical and topological data analysis of real observational data may thus lead to a unique means of studying the phase-space dynamics and singularity structure of the cosmic web. This will yield unique insights in the dynamics of the cosmic web and culminate in a dynamical characterization of the observed distribution of galaxies in upcoming large galaxy surveys such as Euclid and SKA. The potential for additional applications is large, such as the issue of astronomical object classification in data coming from astronomical instruments. In this context, the PhD student will interact with other COFUND PhD students who need advanced methods for understanding the shape of their data.
Keywords: Topological data analysis, Cosmology, Astronomy, Big Data
Fields of expertise involved: Algebraic Topology, Computational Geometry, Cosmology, Astronomy
- Prof. dr. Gert Vegter (JBI)
- Prof. dr. Rien van de Weijgaert (Kapteyn)
- Dr. Konstantinos Efstathiou (JBI)
PhD student: Georg Wilding (Austria)
Potential partners: CIT
Project: A Visual Analytics Approach of Big Data
This project focuses on a Visual Analytics approach of big data, that is, combining automated data analysis techniques with interactive visual interfaces to pose, refine, and confirm hypotheses about complex phenomena represented by such data.
We will focus on high-dimensional time-dependent data, that is, large sets of observations having each many measurement values that represent the evolution of a phenomenon over time. Challenges are here the large numbers of observations (millions or more); dimensions (hundreds or more); and time steps (thousands or more). Finding efficient and effective ways to discover and display complex patterns in this very high dimensional data space is the key challenge in modern (visual) data exploration.
To solve this, we will develop methods for data-size and data-dimensionality reduction; automatic and user-assisted discovery of meaningful patterns hidden in the data; and intuitive visual depiction of such patterns, their evolution, and their inter-relationships. For this, we will develop new techniques for pattern mining, dimensionality reduction, scalable information visualization, uncertainty visualization, relational visualization, and interactive data querying. We will achieve scalability for big data by using CPU and GPU parallelization, and multiscale data-representation and visualization techniques.
We will apply our interactive visual analytics pipeline to two use cases: 1) large simulations or observation catalogs in an astronomical pilot project to address questions on galaxy evolution; 2) prediction of neurodegenerative diseases from multi-centre clinical brain data.
This PhD student follows up on a recently completed project on e-Visualisation of Big Data funded by the Dutch e-Science Centre NLeSC.
Keywords: Visual analytics, high-dimensional data visualization, interactive pattern discovery, scalable information visualization, astronomical data, medical data.
Fields of expertise involved: Multidimensional data analysis, Data and pattern mining, Interactive visual analytics and information visualization, Disease prediction from medical data, Galaxy evolution.
PhD student: Youngjoo Kim (Korea)
Potential partners: Philips
COMPLEX SYSTEMS & ENGINEERING
Project: Distributed control methods under communication constraints with applications to sensor networks in a smart industry setting
In modern industry, there is an increasing interest in control systems employing multiple sensors and actuators, possibly geographically distributed. Communication is an important component of these networked control systems. Understanding the interactions between control and communication components is especially important to develop systems that possess scalability features. Existing solutions for distributed control have little to no scalability features in terms of both convergence time and accuracy caused by limited bandwidth and quantization issues. The ESRs who select this Project line will develop novel distributed control methods, with emphasis on event-based and self-triggered communication protocols. A specific case study will be considered in terms of distributed “average” and “max-min” consensus networks, which are prototypical networks for distributed sensing and actuation.
Keywords: Sensor Networks, Nonlinear consensus, Hybrid Systems, Self-triggered Algorithms, Coordination Control.
Fields of expertise: Control Theory, Dynamical Systems, Signal Processing, Computer Science, Network Science.
PhD student: Monica Rotulo (Italy)
Potential partners: MathSys
Project: Cyber security with applications to both smart energy systems and smart industry
Owing to advances in computing and communication technologies, recent years have witnessed a growing interest towards Cyber-physical Systems (CPSs), i.e., systems where physical processes are monitored/controlled via embedded computers and networks. The concept of CPSs is extremely appealing for smart energy systems and smart industry, but it raises many theoretical and practical challenges. In particular, CPSs have triggered the attention towards networked control in the presence of cyber-attacks. In fact, unlike general-purpose computing systems where attacks limit their impact to the cyber realm, attacks to CPSs can impact the physical world as well. The ESRs who select this Project line will develop novel monitoring and control systems that are resilient against Bias Injection (BI) and Denial-of-Service (DoS) attacks. Two specific case studies will be considered: (i) BI attacks on distributed consensus networks for environmental monitoring, and (ii) DoS attacks on distributed control algorithms for optimal frequency regulation in smart grids.
Keywords: Cyberphysical Systems, Networked Control Systems, Hybrid Systems, Power Networks, Smart Industries.
Fields of expertise: Control Theory, Dynamical Systems, Cybersecurity, Computer Science, Network Science.
PhD student: Alessandro Luppi (Italy)
Potential partners: MathSys
Project: Home robotics
Robots must be able to handle unforeseen circumstances. Neither knowledge representation nor machine learning approaches allow for the sufficiently robust handling of unforeseen circumstances. As a result, new hybrid technology must be developed that combines knowledge technology for the manual representation of behavior-guiding scenarios for new and exceptional circumstances with data technology to evaluate and adapt these scenarios. In the robot architecture developed in the project, a hypothesis testing cycle will be modeled using argumentation-based techniques designed for the combination of logic-based scenario representations and probability-based data analysis. The architecture will be tested in the international annual RoboCup@Home competition.
Keywords: Artificial intelligence, argumentation, robotics, scenario modeling, human-machine interaction.
Fields of expertise involved: Artificial intelligence, argumentation theory, robotics, human-machine interaction.Supervisory team:
Project: Tailor-made model reduction methods for integrated energy systems
The rapid increase of renewable/sustainable energy production leads to daunting challenges in the design, analysis, and control of the modern energy system. Namely, the large number and decentralized nature of renewable energy sources calls for a change of the current paradigm in which energy production is largely centralized in a small number of power plants.
Instead, future energy systems will constitute a network of a large number of heterogeneous dynamical systems, corresponding to a variety of agents such as large power plants, wind farms, solar collectors, industrial and household consumers.
One major challenge in the design of these highly integrated energy systems is that the complexity of the resulting mathematical models is immense, such that the analysis, simulation, and controller design become intractable. Therefore, satisfactory methods for the approximation of such complex models by lower-order, less complex models are needed. In order for these low-complexity models to be reliably employed as substitute of the original model, they should not only accurately preserve the behavior of the original model, also inherit/preserve/reflect structural properties of the original model, in particular the underlying physical network structure. Consequently, this project will investigate network-structure preserving model reduction methods as well as a priori error bounds for such methods.
Keywords: Networks of dynamical systems, structure preserving model reduction, smart energy systems, model reduction for networks, graph simplification
Fields of expertise involved: Systems and control engineering, Feedback control theory, Mathematical modelling, Model reduction, Graph theory.
PhD student: Azka Burohman (Indonesia)
- Centre for Doctoral Training in Mathematics for Real-World Systems (MathSys CDT)
- Centre for information technology (CIT)
ADVANCED INSTRUMENTATION & BIG DATA
Project: Modeling and control of hysteretic deformable mirrors for high-contrast imaging systems
In this project, we will develop model-based nonlinear control algorithm for the control of a novel hysteretic deformable mirror (HDM). For enabling wavefront control in a high-contrast coronagraph instrument for future space-telescopes, with the ultimate goal of finding and characterizing Earth-like exoplanets, we have developed a novel deformable mirror (DM) concept based on new hysteretic piezoelectric material and new distributed polarization method. It allows for a high-density, low-power and scalable DM system which is crucial for space application. The demonstrator is currently being built and tested by a multidisciplinary team from ENTEG (Engineering and Technology institute Groningen), ZIAM (Zernike Institute for Advanced Material), SRON (Dutch Space Research Agency) and KAI (Kapteyn Astronomical Institute).
As an instrumental element of this novel DM system, the smart control system will be developed by the PhD student/Early Stage Researcher (PhD student) within DSSC. In particular, the PhD student will (i). model and characterize the realized HDM demonstrator, including, hysteresis and dynamical behavior; (ii). perform systems identification/identify parameters based on the real-time dynamical data; (iii). develop a distributed control method for achieving desired shape of the DM; and (iv). develop a combination of data-driven and model-based distributed control algorithm for continuous and iterative improvement of the control systems.
Keywords: Control algorithm; Distributed nonlinear control method; Advanced instrumentation; Mechatronic systems; Systems identification; Model-based control algorithm; Data-driven control algorithm.Field of expertise: Systems and Control; Applied Mathematics; Applied Physics; Astronomy; Mechanical Engineering; Electrical & Electronics Engineering; Mechatronics; Advanced Instrumentation; Opto-mechatronics; Precision Engineering.
- Prof. dr. Bayu Jayawardhana (ENTEG)
- Prof. dr. Reynier Peletier (Kapteyn)
- External advisors: Dr. Pieter Dieleman (SRON), Dr. Robert Huisman (SRON), Prof. dr. Beatriz Noheda (ZIAM)
PhD student: Marco Augusto Vasquez Beltran (Mexico)
Project: Scalable algorithms to process massive datasets from radio astronomy
One of the holy grails in cosmology is the discovery of neutral hydrogen in the infant Universe (its first billion years), which can tell astronomers and cosmologists how the first stars, galaxies and black holes formed, how dark matter evolved and whether gravity indeed follows Einstein’s general relativity. The low-frequency part of the Square Kilometer Array (SKA) – the world’s largest radio telescope currently under construction in Australia and South Africa – can not only discover the emission of this hydrogen, but it can even make images of its distribution and how it evolves in time (redshift). The data volume of the SKA, however, is huge (1000 Tb/day) and will require novel scalable algorithms in order to process the total amount of data accumulated in such a project (106 Tb or 1 Exabyte).
The PhD student project will focus on how subtle effects in the processing (i.e. error-correction and imaging) of these data can be uncovered using machine learning techniques, such as neural networks and pattern recognition. Such an approach is becoming necessary in order to quickly explore the very high-dimensional data-set, beyond any human ability, and discover effects and relations in the data that will enable finding these extremely faint signals of neutral hydrogen from the infant Universe.
Keywords: Cosmology, Radio Telescopes, Machine Learning, Infant Universe.
Fields of expertise involved: Cosmic Dawn & Reionization, Radio astronomy, Machine Learning, Signal Processing.
- Prof. dr. Leon Koopmans (Kapteyn)
- Dr. Michael Wilkinson (JBI)
- A third suitable supervisor/advisor might be considered at a later stage, as the project develops, in particular from ASTRON.
PhD student: Hyoyin Gan (Korea)
Project: Methods for automated and robust analysis of astronomical data
This project (Kapteyn, JBI and ASTRON) will focus on methods for automated and robust analysis of astronomical data taken with the Low Frequency Array (LOFAR). The Early Stage Researcher will develop sophisticated calibration methods that connect astronomy and complex data systems, as part of a project on Advanced Instrumentation and Big Data. They will use novel computing science methods to analyse data from the LOFAR radio interferometer, built and operated by ASTRON. LOFAR is the world’s largest connected radio telescope that is optimised to operate a low radio frequencies at arc-second angular resolution. The student will develop new techniques for the automated and robust analysis of the data taken with the International LOFAR Telescope, which includes signals from the stations both within the Netherlands and throughout Europe. This data stream will be used to investigate the low energy radiation from supermassive black holes at the highest possible angular resolution to investigate particle acceleration and energetics in the Universe’s largest natural particle accelerators. This will be done by combining the resolving power of LOFAR and the added magnification provided by gravitational lensing to study black hole physics at redshifts where the active phase from such objects is expected to peak, and have the largest influence on the build-up of the stellar host galaxy. The computing science will involve developing automated methods for identifying such objects from the LOFAR all-sky surveys data. The novel analysis tools developed in this project line will be made available to the community.
Key Words: Galaxy formation; Gravitational lensing; Black Holes
Fields of expertise involved: Radio Astronomy; Interferometry; Advanced calibration; Machine Learning
PhD student: Samira Rezaei (Iran)
Potential partners: Netherlands Institute for Radio Astronomy (ASTRON)
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No. 754315.
|Last modified:||16 August 2019 09.51 a.m.|