Search for: All records

Abstract The COVID-19 outbreak is a global pandemic declared by the World Health Organization, with rapidly increasing cases in most countries. A wide range of research is urgently needed for understanding the COVID-19 pandemic, such as transmissibility, geographic spreading, risk factors for infections, and economic impacts. Reliable data archive and sharing are essential to jump-start innovative research to combat COVID-19. This research is a collaborative and innovative effort in building such an archive, including the collection of various data resources relevant to COVID-19 research, such as daily cases, social media, population mobility, health facilities, climate, socioeconomic data, research articles, policy and regulation, and global news. Due to the heterogeneity between data sources, our effort also includes processing and integrating different datasets based on GIS (Geographic Information System) base maps to make them relatable and comparable. To keep the data files permanent, we published all open data to the Harvard Dataverse ( https://dataverse.harvard.edu/dataverse/2019ncov ), an online data management and sharing platform with a permanent Digital Object Identifier number for each dataset. Finally, preliminary studies are conducted based on the shared COVID-19 datasets and revealed different spatial transmission patterns among mainland China, Italy, and the United States. 

We propose methods for distributed graph-based multi-task learning that are based on weighted averaging of messages from other machines. Uniform averaging or diminishing stepsize in these methods would yield consensus (single task) learning. We show how simply skewing the averaging weights or controlling the stepsize allows learning different, but related, tasks on the different machines. 

We suggest a general oracle-based framework that captures different parallel stochastic optimization settings described by a dependency graph, and derive generic lower bounds in terms of this graph. We then use the framework and derive lower bounds for several specific parallel optimization settings, including delayed updates and parallel processing with intermittent communication. We highlight gaps between lower and upper bounds on the oracle complexity, and cases where the “natural” algorithms are not known to be optimal. 

We study the problem of distributed multitask learning with shared representation, where each machine aims to learn a separate, but related, task in an unknown shared low-dimensional subspaces, i.e. when the predictor matrix has low rank. We consider a setting where each task is handled by a different machine, with samples for the task available locally on the machine, and study communication-efficient methods for exploiting the shared structure.