| 1. |
- Carlsson, Fredrik, et al.
(author)
-
Fine-Grained Controllable Text Generation Using Non-Residual Prompting
- 2022
-
In: Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics. - Stroudsburg, PA, USA : Association for Computational Linguistics. - 9781955917216 ; , s. 6837-6857
-
Conference paper (peer-reviewed)abstract
- The introduction of immensely large Causal Language Models (CLMs) has rejuvenated the interest in open-ended text generation. However, controlling the generative process for these Transformer-based models is at large an unsolved problem. Earlier work has explored either plug-and-play decoding strategies, or more powerful but blunt approaches such as prompting. There hence currently exists a trade-off between fine-grained control, and the capability for more expressive high-level instructions. To alleviate this trade-off, we propose an encoder-decoder architecture that enables intermediate text prompts at arbitrary time steps. We propose a resource-efficient method for converting a pre-trained CLM into this architecture, and demonstrate its potential on various experiments, including the novel task of contextualized word inclusion. Our method provides strong results on multiple experimental settings, proving itself to be both expressive and versatile.
|
|
| 2. |
- Ji, Jiadong, et al.
(author)
-
Detection for pathway effect contributing to disease in systems epidemiology with a case-control design
- 2015
-
In: BMJ Open. - : BMJ. - 2044-6055. ; 5:1, s. e006721-
-
Journal article (peer-reviewed)abstract
- Objectives: Identification of pathway effects responsible for specific diseases has been one of the essential tasks in systems epidemiology. Despite some advance in procedures for distinguishing specific pathway (or network) topology between different disease status, statistical inference at a population level remains unsolved and further development is still needed. To identify the specific pathways contributing to diseases, we attempt to develop powerful statistics which can capture the complex relationship among risk factors. Setting and participants: Acute myeloid leukaemia (AML) data obtained from 133 adults (98 patients and 35 controls; 47% female). Results: Simulation studies indicated that the proposed Pathway Effect Measures (PEM) were stable; bootstrap-based methods outperformed the others, with bias-corrected bootstrap CI method having the highest power. Application to real data of AML successfully identified the specific pathway (Treg -> TGF beta -> Th17) effect contributing to AML with p values less than 0.05 under various methods and the bias-corrected bootstrap CI (-0.214 to -0.020). It demonstrated that Th17-Treg correlation balance was impaired in patients with AML, suggesting that Th17-Treg imbalance potentially plays a role in the pathogenesis of AML. Conclusions: The proposed bootstrap-based PEM are valid and powerful for detecting the specific pathway effect contributing to disease, thus potentially providing new insight into the underlying mechanisms and ways to study the disease effects of specific pathways more comprehensively.
|
|
| 3. |
- Paluch, Amanda E., et al.
(author)
-
Daily steps and all-cause mortality : a meta-analysis of 15 international cohorts
- 2022
-
In: The Lancet Public Health. - : Elsevier. - 2468-2667. ; 7:3, s. e219-e228
-
Journal article (peer-reviewed)abstract
- BACKGROUND: Although 10 000 steps per day is widely promoted to have health benefits, there is little evidence to support this recommendation. We aimed to determine the association between number of steps per day and stepping rate with all-cause mortality.METHODS: In this meta-analysis, we identified studies investigating the effect of daily step count on all-cause mortality in adults (aged ≥18 years), via a previously published systematic review and expert knowledge of the field. We asked participating study investigators to process their participant-level data following a standardised protocol. The primary outcome was all-cause mortality collected from death certificates and country registries. We analysed the dose-response association of steps per day and stepping rate with all-cause mortality. We did Cox proportional hazards regression analyses using study-specific quartiles of steps per day and calculated hazard ratios (HRs) with inverse-variance weighted random effects models.FINDINGS: We identified 15 studies, of which seven were published and eight were unpublished, with study start dates between 1999 and 2018. The total sample included 47 471 adults, among whom there were 3013 deaths (10·1 per 1000 participant-years) over a median follow-up of 7·1 years ([IQR 4·3-9·9]; total sum of follow-up across studies was 297 837 person-years). Quartile median steps per day were 3553 for quartile 1, 5801 for quartile 2, 7842 for quartile 3, and 10 901 for quartile 4. Compared with the lowest quartile, the adjusted HR for all-cause mortality was 0·60 (95% CI 0·51-0·71) for quartile 2, 0·55 (0·49-0·62) for quartile 3, and 0·47 (0·39-0·57) for quartile 4. Restricted cubic splines showed progressively decreasing risk of mortality among adults aged 60 years and older with increasing number of steps per day until 6000-8000 steps per day and among adults younger than 60 years until 8000-10 000 steps per day. Adjusting for number of steps per day, comparing quartile 1 with quartile 4, the association between higher stepping rates and mortality was attenuated but remained significant for a peak of 30 min (HR 0·67 [95% CI 0·56-0·83]) and a peak of 60 min (0·67 [0·50-0·90]), but not significant for time (min per day) spent walking at 40 steps per min or faster (1·12 [0·96-1·32]) and 100 steps per min or faster (0·86 [0·58-1·28]).INTERPRETATION: Taking more steps per day was associated with a progressively lower risk of all-cause mortality, up to a level that varied by age. The findings from this meta-analysis can be used to inform step guidelines for public health promotion of physical activity.FUNDING: US Centers for Disease Control and Prevention.
|
|
| 4. |
- Yang, Jianming, et al.
(author)
-
Comprehensive understanding of heat-induced degradation of triple-cation mixed halide perovskite for a robust solar cell
- 2018
-
In: Nano Energy. - : ELSEVIER SCIENCE BV. - 2211-2855 .- 2211-3282. ; 54, s. 218-226
-
Journal article (peer-reviewed)abstract
- The triple-cation mixed halide perovskite Cs-0.05(MA(0.17)FA(0.83))(0.95)Pb(I0.83Br0.17)(3) emerges as one of the most promising candidates for photovoltaics due to superior optoelectronic properties, but the thermal stability is still a major challenge for the viability of perovskite solar cells towards commercialization. Herein, we firstly explore the thermal response of the photovoltaic performances to access device physical changes. It is shown that the efficiency loss originates from decreased charge mobility, increased trap density and generation of PbI2 charge recombination centers near the interface. In-depth analysis of evolutions in morphology, chemical composition, dynamic and electronic structure of the perovskite layer at the nanometer scales indicates that it is initial dangling bonds and vacancies on the imperfect surfaces decrease the activation energy and cause the perovskite decomposition in a layer-by-layer pathway sequentially from the film surface to bulk. Based on the results, a strategy of surface passivation to improve the thermal stability is demonstrated and discussed. This work for the first time provides insights into the physical and chemical change of such triple-cation perovskite and indicates that more effort should be invested in surface treatment for enhancing perovskite device stability.
|
|
