. Author manuscript; available in PMC: 2019 Sep 1.

Published in final edited form as: IEEE J Biomed Health Inform. 2017 Oct 27;22(5):1589–1604. doi: 10.1109/JBHI.2017.2767063

Table V. Interpretability Techniques for Deep EHR Systems.

Type	Methods
(1) Maximum activation	Convolutional filter response [19]
	Output activation maximization [22]
	Dense top-layer weight maximization [44]
(2) Constraints	Non-negativity [23]
	Non-negative matrix factorization [22]
	Sparsity [50]
	Ontology smoothing [23]
	Regularization [23]
(3) Qualitative clustering	Principal component analysis [49] t-SNE [19]
(4) Mimic learning	Interpretable mimic learning [59]–[61]