Machine learning (ML) is a crucial component of artificial intelligence (AI) and a driving force in our increasingly digital, data-driven world. However, contemporary ML faces significant challenges such as a lack of transparency and explainability, unfairness, insufficient quantification of uncertainty, and high energy requirements. Integreat, the new Norwegian center for knowledge-driven machine learning, contributes to the transformation of ML by developing theories, methods, models, and algorithms that integrate general and domain-specific knowledge with data. By combining statistics, logic, and ML in unique ways, Integreat will produce solutions that are more accurate, sustainable, fair and explainable, and account for uncertainty. Knowledge involves higher-level understanding and representation of facts, mechanisms, relations and structures of the object, system or domain of study, as well as of the data-generating mechanisms and measurement procedures. Knowledge can be “hard” and represented using deterministic frameworks, such as logical formulas, ontologies, causal diagrams, and hierarchical models and relations; or knowledge can be “soft”, represented using stochastic and probable mechanisms, such as associations, potential links, probabilistic models, or just expert opinions in textual form. Knowledge in both of these forms is widespread in all areas of science and society, but incorporating it into ML in a direct, transparent and efficient way is extremely challenging. To test and validate their results, Integreat collaborates with scientists and organizations on real-world problems, which will produce lasting benefits for science and society. With world-leading researchers, young talents, and prominent international scientists, Integreat will shape the new field of knowledge-driven ML in Norway. It is a partnership between the Universities of Oslo and Tromsø and Norsk Regnesentral.The centre starts at the end of 2023 and continues for ten years.
Machine learning (ML) is the mathematical and computational engine of Artificial Intelligence (AI), and as such a fundamental force of technological progress in our increasingly digital, data-driven world. Major obstacles however are that contemporary ML is neither explainable nor transparent, is often unfair, lacks robustness, insufficiently quantifies uncertainty, fails to generalise across domains, depends on huge amounts of curated data, and has algorithms requiring excessive energy to run. Integreat, the Norwegian centre for knowledge-driven machine learning, will contribute to the radical transformation of ML. Expanding the data-centric paradigm of ML, Integreat develops theories, methods, models and algorithms that integrate general and domain-specific knowledge with data, laying the foundations of next generation ML. This will be done by combining the mathematical and computational cultures, and the methodologies of statistics, logic and ML in unique ways. Integreat’s knowledge-driven ML will produce solutions which are (1) more accurate, (2) sustainable, (3) fair and explainable, and (4) account for uncertainty. To reach these four objectives, Integreat’s methodology injects knowledge into Bayesian inference, approximation methods, and into integrative, logic-aware, language-based, causal and transfer learning. To test and validate our strong and deep foundational results on real-world problems, we draw on our collaborations with scientists and organisations. This will produce lasting benefits for science, society and the economy. The synergy of world-leading researchers, who have been and are contributing highly influential methods to statistics, logic and ML, many young talents and prominent international scientists, will allow Integreat to shape the new field of knowledge-driven ML. Integreat will contribute to the national effort to make Norway one of the world’s leading countries for ML research.