has the abstract

Tobias Kuhn

2025-05-26T10:00:49.617Z

has the abstract

Tobias Kuhn

2025-05-23T13:16:52.346Z

The majority of economic sectors are transformed by the abundance of data. Smart grids, smart cities, smart health, Industry 4.0 impose to domain experts requirements for data science skills in order to respond to their duties and the challenges of the digital society. Business training or replacing domain experts with computer scientists can be costly, limiting for the diversity in business sectors and can lead to sacrifice of invaluable domain knowledge. This paper illustrates experience and lessons learnt from the design and teaching of a novel cross-disciplinary data science course at a postgraduate level in a top-class university. The course design is approached from the perspectives of the constructivism and transformative learning theory. Students are introduced to a guideline for a group research project they need to deliver, which is used as a pedagogical artifact for students to unfold their data science skills as well as reflect within their team their domain and prior knowledge. In contrast to other related courses, the course content illustrated is designed to be self-contained for students of different discipline. Without assuming certain prior programming skills, students from different discipline are qualified to practice data science with open-source tools at all stages: data manipulation, interactive graphical analysis, plotting, machine learning and big data analytics. Quantitative and qualitative evaluation with interviews outlines invaluable lessons learnt.

Tobias Kuhn

2025-05-05T09:15:59.143Z

Symbolic approaches to Artificial Intelligence (AI) represent things within a domain of knowledge through physical symbols, combine symbols into symbol expressions, and manipulate symbols and symbol expressions through inference processes. While a large part of Data Science relies on statistics and applies statistical approaches to AI, there is an increasing potential for successfully applying symbolic approaches as well. Symbolic representations and symbolic inference are close to human cognitive representations and therefore comprehensible and interpretable; they are widely used to represent data and metadata, and their specific semantic content must be taken into account for analysis of such information; and human communication largely relies on symbols, making symbolic representations a crucial part in the analysis of natural language. Here we discuss the role symbolic representations and inference can play in Data Science, highlight the research challenges from the perspective of the data scientist, and argue that symbolic methods should become a crucial component of the data scientists’ toolbox.

Tobias Kuhn

2025-05-05T09:13:17.554Z

Research on international conflict has mostly focused on explaining events such as the onset or termination of wars, rather than on trying to predict them. Recently, however, forecasts of political phenomena have received growing attention. Predictions of violent events, in particular, have been increasingly accurate using various methods ranging from expert knowledge to quantitative methods and formal modeling. Yet, we know little about the limits of these approaches, even though information about these limits has critical implications for both future research and policy-making. In particular, are our predictive inaccuracies due to limitations of our models, data, or assumptions, in which case improvements should occur incrementally. Or are there aspects of conflicts that will always remain fundamentally unpredictable? After reviewing some of the current approaches to forecasting conflict, I suggest avenues of research that could disentangle the causes of our current predictive failures.

Tobias Kuhn

2025-04-25T11:18:29.878Z

Data science is a young and rapidly expanding field, but one which has already experienced several waves of temporarily-ubiquitous methodological fashions. In this paper we argue that a diversity of ideas and methodologies is crucial for the long term success of the data science community. Towards the goal of a healthy, diverse ecosystem of different statistical models and approaches, we review how ideas spread in the scientific community and the role of incentives in influencing which research ideas scientists pursue. We conclude with suggestions for how universities, research funders and other actors in the data science community can help to maintain a rich, eclectic statistical environment.

Tobias Kuhn

2025-04-25T11:17:39.508Z

Various approaches and systems have been presented in the context of scholarly communication for what has been called semantic publishing. Closer inspection, however, reveals that these approaches are mostly not about publishing semantic representations, as the name seems to suggest. Rather, they take the processes and outcomes of the current narrative-based publishing system for granted and only work with already published papers. This includes approaches involving semantic annotations, semantic interlinking, semantic integration, and semantic discovery, but with the semantics coming into play only after the publication of the original article. While these are interesting and important approaches, they fall short of providing a vision to transcend the current publishing paradigm. We argue here for taking the term semantic publishing literally and work towards a vision of genuine semantic publishing, where computational tools and algorithms can help us with dealing with the wealth of human knowledge by letting researchers capture their research results with formal semantics from the start, as integral components of their publications. We argue that these semantic components should furthermore cover at least the main claims of the work, that they should originate from the authors themselves, and that they should be fine-grained and light-weight for optimized re-usability and minimized publication overhead. This paper is in fact not just advocating our concept, but is itself a genuine semantic publication, thereby demonstrating and illustrating our points.

Tobias Kuhn

2025-04-24T06:17:02.157Z

Computational manipulation of knowledge is an important, and often under-appreciated, aspect of biomedical Data Science. The first Data Science initiative from the US National Institutes of Health was entitled “Big Data to Knowledge (BD2K).” The main emphasis of the more than $200M allocated to that program has been on “Big Data;” the “Knowledge” component has largely been the implicit assumption that the work will lead to new biomedical knowledge. However, there is long-standing and highly productive work in computational knowledge representation and reasoning, and computational processing of knowledge has a role in the world of Data Science. Knowledge-based biomedical Data Science involves the design and implementation of computer systems that act as if they knew about biomedicine. There are many ways in which a computational approach might act as if it knew something: for example, it might be able to answer a natural language question about a biomedical topic, or pass an exam; it might be able to use existing biomedical knowledge to rank or evaluate hypotheses; it might explain or interpret data in light of prior knowledge, either in a Bayesian or other sort of framework. These are all examples of automated reasoning that act on computational representations of knowledge. After a brief survey of existing approaches to knowledge-based data science, this position paper argues that such research is ripe for expansion, and expanded application.

Tobias Kuhn

2025-04-24T06:15:44.650Z

Stable states in complex systems correspond to local minima on the associated potential energy surface. Transitions between these local minima govern the dynamics of such systems. Precisely determining the transition pathways in complex and high-dimensional systems is challenging because these transitions are rare events, and isolating the relevant species in experiments is difficult. Most of the time, the system remains near a local minimum, with rare, large fluctuations leading to transitions between minima. The probability of such transitions decreases exponentially with the height of the energy barrier, making the system's dynamics highly sensitive to the calculated energy barriers. This work aims to formulate the problem of finding the minimum energy barrier between two stable states in the system's state space as a cost-minimization problem. It is proposed to solve this problem using reinforcement learning algorithms. The exploratory nature of reinforcement learning agents enables efficient sampling and determination of the minimum energy barrier for transitions.

Tobias Kuhn

2025-04-24T06:14:13.990Z

has abstract

Barbara Magagna

2023-01-10T14:07:49.629Z