While the Machine Learning (ML) landscape is evolving rapidly, there has been a relative lag in the development of the "learning systems" needed to enable broad adoption. Furthermore, few such systems are designed to support the specialized requirements of scientific ML. Here we present the Data and Learning Hub for science (DLHub), a multi-tenant system that provides both model repository and serving capabilities with a focus on science applications. DLHub addresses two significant shortcomings in current systems. First, its selfservice model repository allows users to share, publish, verify, reproduce, and reuse models, and addresses concerns related to model reproducibility by packaging and distributing models and all constituent components. Second, it implements scalable and low-latency serving capabilities that can leverage parallel and distributed computing resources to democratize access to published models through a simple web interface. Unlike other model serving frameworks, DLHub can store and serve any Python 3compatible model or processing function, plus multiple-function pipelines. We show that relative to other model serving systems including TensorFlow Serving, SageMaker, and Clipper, DLHub provides greater capabilities, comparable performance without memoization and batching, and significantly better performance when the latter two techniques can be employed. We also describe early uses of DLHub for scientific applications.
This research investigates innovation in how film producers use social digital tools to engage consumers, reduce demand uncertainty and respond to the challenge of digital disruption that affects the traditional film value chain. Through three empirical case studies of film production and exploitation, we examine examples of innovation in product, service, distribution, marketing and process, each having important implications at the organizational level. Our findings show that innovations in one area have important implications for other areas, distribution impacting on concepts of product and service, for example. We also show that internal firm micro-process dynamics impact directly on external interactions between the firm, consumers en masse and partner firms. Our research thus lies at the nexus of innovation, social media and uncertainty management, and questions the boundaries found in innovation 'types' or dominant taxonomies in traditional R&D frames. *This work was supported by the Economic and Social Research Council Capacity Building Cluster Grant RES 187-24-0014Introduction Film production is facing increasing challenges caused by declining revenues from DVD and TV rights exploitation. Digital tools, applied in new marketing and distribution models, form innovative strategic responses to major threats to film businesses caused by digital disruption (UKFC, 2010). These interventions, however, occur far earlier in the product life cycle and are undertaken by different parties than has traditionally been the case and can be seen as the active management of consumer demand uncertainty (Miller & Shamsie, 1999;Dempster, 2006). We ask how social digital tools are applied to manage uncertainty in the UK film business and adopt an empirical case study approach to investigate this. In doing so, we address a gap in the literature at the nexus of innovation, social media and uncertainty management in a specific creative industry, film. Whilst Dempster (2006) explores risk and uncertainty management in theatre and Sgourev (2012) deals with risk and innovation in opera, the specific 'spreadable' nature of digital media (Jenkins, Ford & Green, 2012) has not been explored in an innovation context for managing uncertainty in this setting.
Public comments submitted during agency rulemakings can provide rich insight into stakeholders’ viewpoints around contentious political issues but have been largely untapped as a data source by social scientists. This is in part due to the lack of access to comments in machine-readable formats and in part due to the difficulty in analyzing large corpora of textual data. However, new online repositories and analytic methodologies are beginning to open up this trove of data for researchers. Using data from the online portal regulations.gov, we employ probabilistic topic modeling to identify latent themes in a series of regulatory debates about electronic monitoring in the U.S. trucking industry. Our model suggests that different types of commenters use alternative discursive frames in talking about monitoring. Comments submitted by individuals were more likely to place the electronic monitoring debate in the context of broader logistical problems plaguing the industry, such as long wait times at shippers’ terminals, while organizational stakeholders were more likely than individuals to frame their comments in terms of technological standards and language suggesting cost / benefit quantification.
The real estate financial markets are complex supply chains. Understanding their behavior is limited by a lack of data that would capture the richly interconnected networks of financial institutions and complex financial products, e.g., asset backed securities. This lack of transparency is further compounded by limited knowledge of the contractual rules that control the flow of funds from mortgage pools to securities, as well as the financial events that regulate these flows. In this project, we will use the IBM Midas framework and tools to extract entities, relationships, events, contractual rules and risk profiles for financial institutions. Our source of information will be the MBS prospectus documents that are public and are filed with the Securities and Exchange Commission. We will describe the data management needs of the Haas Real Estate and Financial Markets (REFM) Lab and presents some recent REFM analytics that highlight the importance of these markets and the impact on systemic risk. We use excerpts extracted from the prospectus of a mortgage backed security (MBS) to illustrate the information extraction challenges and outline our approach to address these challenges.
This discussion note provides a perspective on valuation studies by a group of PhD students. Based on impressions from the Valuation as Practice workshop at The University of Edinburgh in early 2014 we were inspired by the example of Kjellberg et al. (2013) to debate how we see, understand, and are inspired by the field of valuation studies. It is the hope of the editors that sharing the concerns of early-stage researchers starting out in a field in flux, may be of use to, and perhaps spur, senior contributors to further develop this emerging research landscape. Using the workshop experience as a springboard, we argue that the domain of valuation studies still relies heavily on influences from the study of economics, with a strong emphasis on processes of quantification and calculation. With apparent pragmatism within the field, concern as to what might be lost by this narrower perspective is raised. Additionally, we call for the exploration of the possibility of a common language of valuation, to better define shared features, and identify as well as manage conflicts within the field.
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.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.