Neural-based multi-task learning has been successfully used in many real-world large-scale applications such as recommendation systems. For example, in movie recommendations, beyond providing users movies which they tend to purchase and watch, the system might also optimize for users liking the movies afterwards. With multi-task learning, we aim to build a single model that learns these multiple goals and tasks simultaneously. However, the prediction quality of commonly used multi-task models is often sensitive to the relationships between tasks. It is therefore important to study the modeling tradeo s between task-speci c objectives and inter-task relationships. In this work, we propose a novel multi-task learning approach, Multi-gate Mixture-of-Experts (MMoE), which explicitly learns to model task relationships from data. We adapt the Mixture-of-Experts (MoE) structure to multi-task learning by sharing the expert submodels across all tasks, while also having a gating network trained to optimize each task. To validate our approach on data with di erent levels of task relatedness, we rst apply it to a synthetic dataset where we control the task relatedness. We show that the proposed approach performs better than baseline methods when the tasks are less related. We also show that the MMoE structure results in an additional trainability bene t, depending on di erent levels of randomness in the training data and model initialization. Furthermore, we demonstrate the performance improvements by MMoE on real tasks including a binary classi cation benchmark, and a large-scale content recommendation system at Google.
BackgroundAn outbreak of dengue fever (DF) occurred in Guangdong Province, China in 2013 with the highest number of cases observed within the preceding ten years. DF cases were clustered in the Pearl River Delta economic zone (PRD) in Guangdong Province, which accounted for 99.6% of all cases in Guangdong province in 2013. The main vector in PRD was Aedes albopictus. We investigated the socioeconomic and environmental factors at the township level and explored how the independent variables jointly affect the DF epidemic in the PRD.Methodology/Principal FindingsSix factors associated with the incidence of DF were identified in this project, representing the urbanization, poverty, accessibility and vegetation, and were considered to be core contributors to the occurrence of DF from the perspective of the social economy and the environment. Analyses were performed with Generalized Additive Models (GAM) to fit parametric and non-parametric functions to the relationships between the response and predictors. We used a spline-smooth technique and plotted the predicted against the observed co-variable value. The distribution of DF cases was over-dispersed and fit the negative binomial function better. The effects of all six socioeconomic and environmental variables were found to be significant at the 0.001 level and the model explained 45.1% of the deviance by DF incidence. There was a higher risk of DF infection among people living at the prefectural boundary or in the urban areas than among those living in other areas in the PRD. The relative risk of living at the prefectural boundary was higher than that of living in the urban areas. The associations between the DF cases and population density, GDP per capita, road density, and NDVI were nonlinear. In general, higher “road density” or lower “GDP per capita” were considered to be consistent risk factors. Moreover, higher or lower values of “population density” and “NDVI” could result in an increase in DF cases.ConclusionIn this study, we presented an effect analysis of socioeconomic and environmental factors on DF occurrence at the smallest administrative unit (township level) for the first time in China. GAM was used to effectively detect the nonlinear impact of the predictors on the outcome. The results showed that the relative importance of different risk factors may vary across the PRD. This work improves our understanding of the differences and effects of socioeconomic and environmental factors on DF and supports effectively targeted prevention and control measures.
BackgroundMalaria has re-emerged in Anhui Province, China, and this province was the most seriously affected by malaria during 2005–2006. It is necessary to understand the spatial distribution of malaria cases and to identify highly endemic areas for future public health planning and resource allocation in Anhui Province.MethodsThe annual average incidence at the county level was calculated using malaria cases reported between 2000 and 2006 in Anhui Province. GIS-based spatial analyses were conducted to detect spatial distribution and clustering of malaria incidence at the county level.ResultsThe spatial distribution of malaria cases in Anhui Province from 2000 to 2006 was mapped at the county level to show crude incidence, excess hazard and spatial smoothed incidence. Spatial cluster analysis suggested 10 and 24 counties were at increased risk for malaria (P < 0.001) with the maximum spatial cluster sizes at < 50% and < 25% of the total population, respectively.ConclusionThe application of GIS, together with spatial statistical techniques, provide a means to quantify explicit malaria risks and to further identify environmental factors responsible for the re-emerged malaria risks. Future public health planning and resource allocation in Anhui Province should be focused on the maximum spatial cluster region.
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.