Fig. 5From: TransCode: Uncovering COVID-19 transmission patterns via deep learningTransCode validation. a Data processing procedure. Each row of the constituency-level TransCode matrix of a district is summated to form the exported transmission intensity of the corresponding constituency. b Gaussian process regression with deep kernel. The model uses the historical exported transmission intensity as the input and infers the future transmission risk of each constituency. c Transmission risk prediction results for the district Yau Tsim Mong (E). Darker colors indicate a higher predicted transmission intensity; the constituency Jordan North (E18) has the highest predicted riskBack to article page