Presentation
(WHPC02) High-volume data processing for ambient healthcare research prototyping
SessionWomen in HPC Posters
Event Type
Women in HPC Posters
AI/Machine Learning/Deep Learning
Big Data Analytics
Personalized Medicine
TimeWednesday, June 27th8:30am - 5pm
LocationAreal
DescriptionThe SPHERE project continues to deploy multimodal sensor networks into dozens of homes in the South West of England. The resulting dataset is expected ultimately to be large enough to significantly complicate reprocessing using the methods developed during the prototype phase of SPHERE, which were designed in the expectation that they would be deployed locally to each home. Therefore, various alternative means for processing the data are considered, including distributed indexing execution. For reasons of confidentiality and pragmatism, the use of public cloud platforms is not appropriate. The University of Bristol's HPC facility, Bluecrystal, will be used for this purpose.
We are currently in the process of adapting our existing processing pipelines to the platform. Once this is complete, we expect the increase in processing speed and parallelism to permit us to compare the behaviour of several machine learning approaches across the dataset. Where practical, the most successful methods will eventually be adapted to run on the sensor platforms themselves.
We are currently in the process of adapting our existing processing pipelines to the platform. Once this is complete, we expect the increase in processing speed and parallelism to permit us to compare the behaviour of several machine learning approaches across the dataset. Where practical, the most successful methods will eventually be adapted to run on the sensor platforms themselves.
Poster PDF