Deep ensembles can be seen as the current state-of-the-art for uncertainty quantification in deep learning. While the approach was originally proposed as an non-Bayesian technique, arguments towards its Bayesian footing have been put forward as well. We show that deep ensembles can be viewed as an approximate Bayesian method by specifying the corresponding assumptions. Our finding leads to an improved approximation which results in an increased epistemic part of the uncertainty. Numerical examples suggest that the improved approximation can lead to more reliable uncertainties. Analytical derivations ensure easy calculation of results.
Information and communication technologies enable migrants to maintain bonds with multiple communities. Little is known about the association between migrants’ connections to their country of origin and different integration practices in online and offline communities in the receiving society. We draw on a survey conducted amongst migrants in Iceland (N = 2,139) and conduct three regression analyses to identify determinants of migrants’ use of media and social media from their country of origin. Contrary to other studies, we do not find evidence of reactive transnationalism (i.e., migrants seeking out connections to their places of origin due to dissatisfaction with life in the receiving society) as a response to negative attitudes towards the receiving society. We identify distinct patterns of online and offline integration: Migrants with frequent contact with their countries of origin are less integrated locally in terms of offline activities. However, they are more integrated in digital communities of the receiving society, and use receiving-country media more frequently, thus following a strategy of digital biculturalism.
This article explores migrants' language learning experiences in two small language communities in the West Nordic Region. We provide a comparative perspective based on an online survey and ethnographic interviews conducted in Iceland and qualitative research conducted in the Faroe Islands. A major finding from this study is that investment in language learning is a highly situated type of activity, which is contingent on personal circumstances, and on structural conditions. Prevailing language ideologies, such as purism and authenticity, can pose constraints on the language learning process among learners who are initially motivated to learn the language. Results show that many migrants follow a utilitarian approach to learning and perceived usefulness of languages influences participants' linguistic choices. A lack of opportunities for language learning has been mentioned by learners in both places we investigate.
Uncertainty quantification by ensemble learning is explored in terms of an application from computational optical form measurements. The application requires to solve a large-scale, nonlinear inverse problem. Ensemble learning is used to extend a recently developed deep learning approach for this application in order to provide an uncertainty quantification of its predicted solution to the inverse problem. By systematically inserting out-of-distribution errors as well as noisy data the reliability of the developed uncertainty quantification is explored. Results are encouraging and the proposed application exemplifies the ability of ensemble methods to make trustworthy predictions on high dimensional data in a real-world application.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.