Big Data Cybernetics

In today’s science and technology, the spatial, temporal and property-profile domains are often handled by different academic disciplines. However, real-world systems have spatial extent, temporal dynamics and a variety of physical properties. Modern measurement devices increasingly allow us to link these domains, which can provide us with deeper understanding, better control and new opportunities. However, the rapid increase in the amount of data currently poses a major challenge which requires a corresponding increase in our ability to interpret and make sense of this “big data”. Many approaches to handling big data are based on black-box methods which may not be intuitive or transparent for human interpretation. Using such methods, humans are not included in the data modelling process and thus also risk being pushed out of the data analysis and decision-making processes.

To address this challenge, a new interdisciplinary field called “Big Data Cybernetics” is envisioned, which combines methods from automatic control and multivariate data modelling in order to discover systematic structures in the spatial, temporal and property-profile domains, and to convert these structures into quantitative, human-interpretable information. For example, the two fields of cybernetics and chemometrics have both developed successful, but quite different modelling tools. While the former focuses on modelling and control in the time domain, the latter focuses on multichannel spectroscopy and multivariate calibration, image analysis and soft modelling.

The main goal of the new field is to translate “big data” from a large number of sensor channels into ”smart data” represented by a combination of theory-driven and data-driven models, by combining science’s prior knowledge with nature’s unexpected patterns to identify the relevant structures and develop interpretable and useful models. The overlap between cybernetic subspace identification and chemometric partial-least-squares regression could for instance be a fruitful common ground for the desired high-dimensional, spatio-temporal modelling. The outputs from such models shall be intuitively understandable by humans, who then can use their background knowledge and creativity for further refinement and development. This means that black-box modelling, such as e.g. artificial neural networks or support vector machines, are not the focus of Big Data Cybernetics.

Together with the Norwegian high-tech company KONGSBERG, the Norwegian University of Science and Technology (NTNU) is currently establishing the world’s first professorship in Big Data Cybernetics. The position will be affiliated with the Department of Engineering Cybernetics at NTNU’s Faculty of Information Technology, Mathematics and Electrical Engineering. The announcement can be found here, with application deadline 30.09.2016.

For the successful applicant, this represents a unique opportunity to play a central role in the development of a new interdisciplinary field. In particular, the new professor will work to establish a cross-disciplinary culture and methodology for Big Data Cybernetics, combining the best of three science and technology traditions: Theory-driven mechanistic modelling, practice-driven multivariate data modelling and advanced control theory. By interpreting the essential patterns found in the data and comparing these to prior mechanistic theory, it should be possible to attain improved mathematical models of the system at hand and better human understanding, thus leading to new scientific discovery.

The new professor is expected to establish and teach university courses, organize seminar series, develop teaching material and spearhead an interdisciplinary scientific network and possibly also a suitable international conference series in Big Data Cybernetics, in close collaboration with colleagues at the department.

It is required to document solid competence in at least one of the two fields of automatic control and multivariate data modelling, and applicants must demonstrate a strong interest in merging these two fields. Knowledge in system identification, nonlinear dynamics, feedback control and self-organization, signal processing, image analysis, visualization or machine learning is an advantage.

Relevant terms related to Big Data Cybernetics:

• Measurement data. Control of dynamic systems usually calls for more or less continuously measured data flows. Using cheap, numerous and multichannel data-sensing mobile devices, aerial (remote sensing), software logs, cameras, microphones, radio-frequency identification (RFID) readers and wireless sensor networks, the capacity to generate informative data sets is growing rapidly.
• Big data is a term for data sets that are so large or complex that traditional data processing applications are inadequate. In 2012, Gartner updated its definition as follows: "Big data is high volume, high velocity, and/or high variety information assets that require new forms of processing to enable enhanced decision making, insight discovery and process optimization." Challenges in Big Data include analysis, capture, data curation, search, sharing, storage, transfer, visualization, querying, updating and information privacy. The term Big Data often refers simply to the use of predictive analytics or certain other advanced methods to extract value from data, and seldom to a particular size of data set.
• Cybernetics is a transdisciplinary approach for exploring regulatory systems, their structures, constraints and possibilities. In the 21st century, the term is often used in a rather loose way to imply "control of any system using technology." Cybernetics is relevant to the study of systems, such as mechanical, physical, biological, cognitive, and social systems. Cybernetics is applicable when a system being analyzed incorporates a closed signaling loop; that is, where action by the system generates some change in its environment and that change is reflected in that system in some manner (feedback) that triggers a system change, originally referred to as a "circular causal" relationship. More specifically, cybernetics involves control theory, which is an interdisciplinary subfield of science that originated in engineering and mathematics, and that deals with influencing the behavior of dynamical systems e.g. through feedback control.
• Subspace data modelling. A prerequisite for coming to utilize full-fledged mega-variate sensor inputs in Big Data Cybernetics is for cybernetics to move from conventional, individually-selective sensors (e. g. individual temperature, pressure and position sensors) to multi-channel sensors (thermal cameras, frequency spectra, spectrophotometers, sensor arrays). Converting such high-dimensional, but non-selective input signals into selective input variables requires multivariate calibration and other reduced-rank, subspace-based multivariate data modelling methods, which hold promising potentials for data simplification in Big Data Cybernetics when combined with subspace-identification methods from control theory.
• Chemometrics is the science of extracting information from chemical systems by data-driven means. Chemometrics is inherently interdisciplinary, using methods frequently employed in core data-analytic disciplines such as multivariate statistics, applied mathematics and computer science, in order to address problems in chemistry, biochemistry, medicine, biology and chemical engineering. In this way, chemometrics mirrors other interdisciplinary fields such as psychometrics and econometrics.