Refine
Departments, institutes and facilities
- Fachbereich Angewandte Naturwissenschaften (56)
- Fachbereich Wirtschaftswissenschaften (33)
- Fachbereich Informatik (31)
- Institut für funktionale Gen-Analytik (IFGA) (25)
- Fachbereich Ingenieurwissenschaften und Kommunikation (19)
- Fachbereich Sozialpolitik und Soziale Sicherung (18)
- Institut für Technik, Ressourcenschonung und Energieeffizienz (TREE) (16)
- Sprachenzentrum (10)
- Institut für Verbraucherinformatik (IVI) (9)
- Internationales Zentrum für Nachhaltige Entwicklung (IZNE) (7)
Document Type
- Part of a Book (249) (remove)
Year of publication
Language
- English (249) (remove)
Keywords
- GC/MS (5)
- Polymers (4)
- Lignin (3)
- Mesenchymal stem cells (3)
- Pyrolysis (3)
- Regenerative medicine (3)
- Scaffolds (3)
- unfolded protein response (3)
- Analytical pyrolysis (2)
- Antimicrobial activity (2)
Deployment of modern data-driven machine learning methods, most often realized by deep neural networks (DNNs), in safety-critical applications such as health care, industrial plant control, or autonomous driving is highly challenging due to numerous model-inherent shortcomings. These shortcomings are diverse and range from a lack of generalization over insufficient interpretability and implausible predictions to directed attacks by means of malicious inputs. Cyber-physical systems employing DNNs are therefore likely to suffer from so-called safety concerns, properties that preclude their deployment as no argument or experimental setup can help to assess the remaining risk. In recent years, an abundance of state-of-the-art techniques aiming to address these safety concerns has emerged. This chapter provides a structured and broad overview of them. We first identify categories of insufficiencies to then describe research activities aiming at their detection, quantification, or mitigation. Our work addresses machine learning experts and safety engineers alike: The former ones might profit from the broad range of machine learning topics covered and discussions on limitations of recent methods. The latter ones might gain insights into the specifics of modern machine learning methods. We hope that this contribution fuels discussions on desiderata for machine learning systems and strategies on how to help to advance existing approaches accordingly.
From Conclusion to Coda
(2022)
Differential-Algebraic Equations and Beyond: From Smooth to Nonsmooth Constrained Dynamical Systems
(2022)
Virtual exchange
(2022)
Orešković and Porsdam Mann draw a distinction between ‘fast’ and ‘slow’ science. Whereas the latter involves rigorous and laborious adherence to the scientific method, the former represents the reality that much scientific work faces time pressures which at times force shortcuts. The distinction can be seen to operate in contemporary research into the coronavirus pandemic: whereas the development of vaccines and treatments usually requires years of meticulous laboratory work and several more years of clinical testing, the many millions suffering from the disease need a treatment now. However, by taking too many safeguards off the treatment discovery and testing pipelines, or by refusing to act in accordance with scientific advice, governments risk sacrificing the public’s trust not only in the government’s scientific bona fides but in the scientific process itself. This is a heavy price to pay, argue Orešković and Porsdam Mann, and point to evidence indicating that the success of Germany and Japan in combating COVID-19 can be traced to public trust in science and government, as well as scientifically-informed and respectful national leadership.