Periodic Reporting for period 2 - DSSC (Data Science and Systems Complexity Research Training Programme)
Période du rapport: 2019-05-01 au 2022-12-31
Problem:
Today’s world is a highly complex system that produces massive amounts of data daily. We can only progress towards urgent global priorities such as sustainable energy, faultless and smart industry or the data-driven economy if we can take into account the dynamic nature of this complex system. In dealing with this challenge, data and complexity research has so far followed separate disciplinary lines. However, such an approach risks to overlook an important aspect of Big Data and complexity, i.e. that complex systems generate big data, and big data help identify, control and analyze complex systems.
The Centre for Data Science and Systems Complexity (DSSC) at the University of Groningen and its COFUND Doctoral Programme address this scientific problem by combining data and complexity research to understand and control complex systems and processes through massive data. The DSSC adopts an interdisciplinary approach and addresses data science and complexity questions from disciplines that range from the fundamental to the applied (mathematics, computer science, artificial intelligence, engineering, astronomy, physics). How to store, archive, and process massive amounts of digital information? How do the parts of a complex system give rise to collective emergent behaviour of the system? How does the system interact, adapt, and how can it be controlled?
Societal importance:
There is a great need for taking into account the dynamic nature of a complex system as today’s world demands. An integrated approach is required that uses data to understand the behaviour of complex systems, by educating experts who can handle data and complexity problems for society at large. In response to this problem, the DSSC innovative doctoral programme trains 10 early stage researchers to become interdisciplinary specialists skilled in both Data and Complexity science, with international, and intersectoral experience, who can solve for Europe the problems associated with the data avalanche and systems complexity.
Overall objectives
The first objective at the DSSC is to contribute to the societal demand for data and complexity experts by training interdisciplinary specialists and innovators in data and complexity science, endowed with management skills and well embedded in networks in academia and the private sector. The second objective is to help solve current scientific and societal problems related to data and complexity by:
- developing trained, generic algorithms that can overcome the fragmentation of domain expertise and solve families of problems
- developing methods to determine the appropriate amounts of data/computing from a variety of sources to control complex systems with the aim to apply just-enough-data
- creating new methods to design algorithms for monitoring and control, which work in the presence of highly uncertain models or even in the absence of them
- demonstrating these new algorithms’ effectiveness on real complex systems, e.g. energy systems, high-tech industry, etc.
More information about our training project and results is available at https://www.rug.nl/research/fse/themes/dssc/cofund/(s’ouvre dans une nouvelle fenêtre)
Today’s world is a highly complex system that produces massive amounts of data daily. We can only progress towards urgent global priorities such as sustainable energy, faultless and smart industry or the data-driven economy if we can take into account the dynamic nature of this complex system. In dealing with this challenge, data and complexity research has so far followed separate disciplinary lines. However, such an approach risks to overlook an important aspect of Big Data and complexity, i.e. that complex systems generate big data, and big data help identify, control and analyze complex systems.
The Centre for Data Science and Systems Complexity (DSSC) at the University of Groningen and its COFUND Doctoral Programme address this scientific problem by combining data and complexity research to understand and control complex systems and processes through massive data. The DSSC adopts an interdisciplinary approach and addresses data science and complexity questions from disciplines that range from the fundamental to the applied (mathematics, computer science, artificial intelligence, engineering, astronomy, physics). How to store, archive, and process massive amounts of digital information? How do the parts of a complex system give rise to collective emergent behaviour of the system? How does the system interact, adapt, and how can it be controlled?
Societal importance:
There is a great need for taking into account the dynamic nature of a complex system as today’s world demands. An integrated approach is required that uses data to understand the behaviour of complex systems, by educating experts who can handle data and complexity problems for society at large. In response to this problem, the DSSC innovative doctoral programme trains 10 early stage researchers to become interdisciplinary specialists skilled in both Data and Complexity science, with international, and intersectoral experience, who can solve for Europe the problems associated with the data avalanche and systems complexity.
Overall objectives
The first objective at the DSSC is to contribute to the societal demand for data and complexity experts by training interdisciplinary specialists and innovators in data and complexity science, endowed with management skills and well embedded in networks in academia and the private sector. The second objective is to help solve current scientific and societal problems related to data and complexity by:
- developing trained, generic algorithms that can overcome the fragmentation of domain expertise and solve families of problems
- developing methods to determine the appropriate amounts of data/computing from a variety of sources to control complex systems with the aim to apply just-enough-data
- creating new methods to design algorithms for monitoring and control, which work in the presence of highly uncertain models or even in the absence of them
- demonstrating these new algorithms’ effectiveness on real complex systems, e.g. energy systems, high-tech industry, etc.
More information about our training project and results is available at https://www.rug.nl/research/fse/themes/dssc/cofund/(s’ouvre dans une nouvelle fenêtre)
The work towards these objectives consisted of training & research activities carried out by the research teams consisting of DSSC supervisors and ESRs, which were supported by 4 administrative work packages. These were: Management (consisting of a steering committee that monitored the progress of the project and research), Evaluation and Selection (consisting of the organization of the call and the establishment of an impartial selection committee of specialists), Dissemination of the programme and its calls (both for the popularization of our call for applications and for our research results in general), and Ethics issues (for the encouragement of an ethical behavior among our researchers). The administrative work packages took place as scheduled and intended. The DSSC received 120 applications from around the world for 10 ESR positions that were all occupied after the first call for applications.
The training for the DSSC COFUND ESRs was decided jointly by the ESRs and their supervisors and included. The research activities were conducted by the 10 ESRs together with their supervisors at the institutes participating in the DSSC. The ESRs had regular research meetings with their supervisors, with each other during bi-annual PhD meetings, and with their research colleagues at the institutes where they were embedded. Their results were disseminated to the scientific community through conference papers and posters and publications (the ESRs currently have more than 30 published articles and conference proceedings).
The training for the DSSC COFUND ESRs was decided jointly by the ESRs and their supervisors and included. The research activities were conducted by the 10 ESRs together with their supervisors at the institutes participating in the DSSC. The ESRs had regular research meetings with their supervisors, with each other during bi-annual PhD meetings, and with their research colleagues at the institutes where they were embedded. Their results were disseminated to the scientific community through conference papers and posters and publications (the ESRs currently have more than 30 published articles and conference proceedings).
One of the main forms of impact of the DSSC project is that it has helped the ESRs to start building their scientific reputation and careers. Currently, all of our ESRs are cited in the academic community and their findings are adopted by their peers. All the ESRs who graduated from the programme have begun new positions in Europe and abroad, both in academia and in the private sector.
At an organisational level, we are also proud that our model of interdisciplinary research and training was adopted at a greater scale through the establishment of a new initiative, the Jantina Tammes School for Digital Society, Technology, and Artificial Intelligence that encourages interdisciplinary projects in areas including Big Data, which are developed on a larger scale at the University of Groningen and which connect to society.
In what concerns our research results, they consist of:
- several new algorithms for the computation of so-called alpha trees, both sequential and shared-memory parallel, suitable for extreme dynamic range data;
- novel methods for visual exploration of large and multidimensional datasets for pattern finding;
- codes built on Distributed Connected Component Filtering and Analysis, which have been further expanded to include self-organising map (SOM) codes and codes that allow features in the SOMs to be selected and visualised.
- several algorithms for the automatic design of controllers from data, which solve problems like - to name a few - the online learning of controllers for switched systems and the data-driven design of nonlinear and safe controllers;
- new techniques for the problems of open-ended learning, online incremental learning, and explainable learning;
- tailor-made model reduction techniques for integrated energy systems, in particular model reduction techniques even if the data are not rich enough to identify the underlying energy system;
- new machine and deep learning algorithms for source detection and characterisation from large astronomical datasets taken at radio wavelengths;
- new paths towards understanding the cosmic web, and exploiting its cosmological information content, which surpass previous attempts and studies that invoked topological data analysis;
- modelling and control methods for a novel deformable mirror technology which utilises special piezoelectric material exhibiting large hysteresis property in order to realize high-density actuator array systems for deformable mirrors.
The publications based on these results are available at https://research.rug.nl/en/organisations/centre-for-data-science-and-systems-complexity-dssc(s’ouvre dans une nouvelle fenêtre)
At an organisational level, we are also proud that our model of interdisciplinary research and training was adopted at a greater scale through the establishment of a new initiative, the Jantina Tammes School for Digital Society, Technology, and Artificial Intelligence that encourages interdisciplinary projects in areas including Big Data, which are developed on a larger scale at the University of Groningen and which connect to society.
In what concerns our research results, they consist of:
- several new algorithms for the computation of so-called alpha trees, both sequential and shared-memory parallel, suitable for extreme dynamic range data;
- novel methods for visual exploration of large and multidimensional datasets for pattern finding;
- codes built on Distributed Connected Component Filtering and Analysis, which have been further expanded to include self-organising map (SOM) codes and codes that allow features in the SOMs to be selected and visualised.
- several algorithms for the automatic design of controllers from data, which solve problems like - to name a few - the online learning of controllers for switched systems and the data-driven design of nonlinear and safe controllers;
- new techniques for the problems of open-ended learning, online incremental learning, and explainable learning;
- tailor-made model reduction techniques for integrated energy systems, in particular model reduction techniques even if the data are not rich enough to identify the underlying energy system;
- new machine and deep learning algorithms for source detection and characterisation from large astronomical datasets taken at radio wavelengths;
- new paths towards understanding the cosmic web, and exploiting its cosmological information content, which surpass previous attempts and studies that invoked topological data analysis;
- modelling and control methods for a novel deformable mirror technology which utilises special piezoelectric material exhibiting large hysteresis property in order to realize high-density actuator array systems for deformable mirrors.
The publications based on these results are available at https://research.rug.nl/en/organisations/centre-for-data-science-and-systems-complexity-dssc(s’ouvre dans une nouvelle fenêtre)