宾夕法尼亚大学

Deep learning has revolutionized modern data science. However, how to accurately quantify the uncertainty of predictions from large-scale deep neural networks (DNNs) remains an unresolved issue. To address this issue, we introduce a novel post-processing approach. This approach feeds the output from the last hidden layer of a pre-trained large-scale DNN model into a stochastic neural network (StoNet), then trains the StoNet with a sparse penalty on a validation dataset and constructs prediction intervals for future observations. We establish a theoretical guarantee for the validity of this approach; in particular, the parameter estimation consistency for the sparse StoNet is essential for the success of this approach. Comprehensive experiments demonstrate that the proposed approach can construct honest confidence intervals with shorter interval lengths compared to conformal methods and achieves better calibration compared to other post-hoc calibration techniques. Additionally, we show that the StoNet formulation provides us with a platform to adapt sparse learning theory and methods from linear models to DNNs.

This paper reviews emerging research on how different interaction modalities influence information processing and behavior when using new technological devices. A conceptual framework of modality effects is proposed which argues that different input (vocal, haptic) and output modalities (audial, visual) can trigger both cognitive and affective processes that alter users' downstream behaviors. For example, information consumed audially from voice technology rather than visually on a screen can enhance both cognitive load (i.e., cognitive process) and one’s sense of social presence (i.e., affective process), which ultimately shape the types of choices made through the technology. The work concludes with a discussion of future research directions in this domain.

Acetoin is a natural compound and is widely used in various industries such as food, dairy, cosmetics, paint, and pharmaceutical industry. The study aimed to enhance acetoin production through Bacillus subtilis by employing strain selection, mutagenesis, nutritional evaluation, and utilization of cell immobilization for enhanced acetoin production. A strain coded as SFS-13 was isolated from agricultural soil and identified as B. subtilis SFS-13. The acetoin production and glucose-consuming efficiency by wild-type SFS-13 were 8.62 ± 0.04 g/L and 25.33 ± 0.02

Mindfulness-based cognitive therapy (MBCT) integrates cognitive therapy and mindfulness meditation to help individuals disengage from maladaptive thoughts and reactive behavioral patterns. By cultivating nonjudgmental present-moment awareness and decentering from thoughts and emotions, MBCT enhances the capacity to respond more skillfully, thereby reducing impulsivity and promoting self-control. This study pilot-tested the effectiveness of MBCT for internet gaming disorder (MBCT-G) among college-aged students at risk for problematic gaming, focusing on reducing symptoms of internet gaming disorder and secondarily on psychological benefits. A pilot experimental study was conducted in South Korea with 46 college-aged students at high risk of internet gaming disorder. The intervention group (n = 24) participated in eight group sessions of MBCT-G and a control group (n = 22) received no intervention. Internet gaming disorder, impulsivity, depressive symptoms, perceived stress, anxiety, and self-control were assessed before and after the intervention and at 4 weeks postintervention. Linear mixed-model analyses indicated participants in the intervention group had greater reductions in internet gaming disorder symptoms, perceived stress, and state anxiety and enhanced improvement in self-control relative to the control group; these changes persisted at 4-week follow-up except for state anxiety. Findings suggested MBCT-G is a promising intervention for college-aged students whose gaming is problematic and puts them at risk for gaming disorder. Further research is needed with a larger sample, a comparison group receiving an alternative intervention and a longer follow-up period. This study was not prospectively registered; however, it was retrospectively registered with ISRCTN Registry (# ISRCTN17598517).

Despite the ethical stance of new energy proponents, solar installations can, like any large-scale investments, be drivers of violent socio-political and socio-economic conflict in fragile contexts. This poses risks both to people and to successful implementation of a green transition. We first review how “green” energy investments can turn “red” through their entanglement in the political economy of conflict. We then empirically analyze the conflict impact of solar projects using a geospatial panel dataset. Employing propensity score panel matching in a subnational research design, we show that commercial-scale solar projects in the 104 countries characterized as fragile contexts at some point between 1994–2020, on average, cause an increase in deaths from conflict in the year following the project start in their local project area. Conflict escalation is driven by sentiment towards elites, underlining the propensity of these projects to undermine social cohesion. Given the ethical deficiency of outcomes that further burden the already poor and vulnerable, these results call for greater accountability for conflict risk assessment and mitigation by green financiers, the companies they fund, green energy advocates, and policy makers together embedded in a too narrow view of the ethical landscape of the green transition.