Skip to main content
. 2020 May 30;2020:98–107.

Figure 2:

Figure 2:

The LSTM network for predicting exacerbations. Time-stamped event variables x 1x n are represented by formulating a sequence of vectors, with each vector representing the events at a given time-stamp x 1,t ... x n,t. Diagnoses and interventions are embedded into dense, lower-dimensional vectors. The static demographic variables x n+1x m feed directly into the output layer of the network.