US 11,790,216 B2
Predicting likelihoods of conditions being satisfied using recurrent neural networks
Gregory Sean Corrado, San Francisco, CA (US); Ilya Sutskever, San Francisco, CA (US); and Jeffrey Adgate Dean, Palo Alto, CA (US)
Assigned to Google LLC, Mountain View, CA (US)
Filed by Google LLC, Mountain View, CA (US)
Filed on Jul. 27, 2020, as Appl. No. 16/940,131.
Application 16/940,131 is a continuation of application No. 15/588,535, filed on May 5, 2017, granted, now 10,726,327.
Application 15/588,535 is a continuation of application No. 15/150,091, filed on May 9, 2016, granted, now 9,646,244, issued on May 9, 2017.
Application 15/150,091 is a continuation of application No. 14/810,381, filed on Jul. 27, 2015, granted, now 9,336,482, issued on May 10, 2016.
Prior Publication US 2021/0019604 A1, Jan. 21, 2021
This patent is subject to a terminal disclaimer.
Int. Cl. G06N 3/08 (2023.01); G06N 3/047 (2023.01); G16H 50/20 (2018.01); G06N 3/042 (2023.01); G06N 3/044 (2023.01); G06N 3/063 (2023.01); G06N 3/02 (2006.01)
CPC G06N 3/047 (2023.01) [G06N 3/042 (2023.01); G06N 3/044 (2023.01); G06N 3/063 (2013.01); G16H 50/20 (2018.01); G06N 3/02 (2013.01); G06N 3/08 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
obtaining a first temporal sequence of health events, wherein the first temporal sequence comprises respective health-related data associated with a particular patient at each of a plurality of time steps;
initializing a respective internal state for each of one or more recurrent neural network layers of a recurrent neural network;
for each of the plurality of time steps, processing the respective health-related data associated with the particular patient at the time step using the recurrent neural network, wherein the processing comprises updating the respective internal state of each of the one or more recurrent neural network layers of the recurrent neural network using the respective health-related data associated with the particular patient at the time step to generate a network internal state of the recurrent neural network for the time step;
generating, from the network internal state of the recurrent neural network after a last time step in the first temporal sequence, a neural network output for the first temporal sequence; and
generating, from the neural network output for the first temporal sequence, health analysis data that characterizes future health events that may occur after a last time step in the first temporal sequence.