Privacy-preserving collaborative machine learning on genomic data using TensorFlow
[article]
Cheng Hong, Zhicong Huang, Wen-jie Lu, Hunter Qu, Li Ma, Morten Dahl, Jason Mancuso
2020
arXiv
pre-print
We design and implement several MPC-friendly ML primitives, including class weight adjustment and parallelizable approximation of activation function. ...
However, genomic data are often held by different stakeholders (e.g. hospitals, universities, and healthcare companies) who consider the data as sensitive information, even though they desire to collaborate ...
Note that we cannot lookup C yi directly since y i is a private input, so instead we compute it via C yi = (C 1 − C 0 ) · y i + C 0 . Parallel piecewise approximation of the sigmoid function. ...
arXiv:2002.04344v2
fatcat:dyiaoxewvfa3pcdcyn5yd2fh7i