US 11,669,723 B2
Data object classification using an optimized neural network
Hayko Jochen Wilhelm Riemenschneider, Zurich (CH); Leonhard Markus Helminger, Zurich (CH); Christopher Richard Schroers, Uster (CH); and Abdelaziz Djelouah, Zurich (CH)
Assigned to Disney Enterprises, Inc., Burbank, CA (US); and ETH Zürich (EIDGENÖSSISCHE TECHNISCHE HOCHSCHULE ZÜRICH), Zurich (CH)
Filed by Disney Enterprises, inc., Burbank, CA (US); and ETH Zürich (EIDGENÖSSISCHE TECHNISCHE HOCHSCHULE ZÜRICH), Zürich (CH)
Filed on Sep. 16, 2022, as Appl. No. 17/946,907.
Application 17/946,907 is a continuation of application No. 16/808,069, filed on Mar. 3, 2020, granted, now 11,475,280.
Claims priority of provisional application 62/936,125, filed on Nov. 15, 2019.
Prior Publication US 2023/0009121 A1, Jan. 12, 2023
Int. Cl. G06N 3/048 (2023.01); G06N 3/08 (2023.01); G06V 10/75 (2022.01); G06F 18/24 (2023.01); G06F 18/214 (2023.01)
CPC G06N 3/048 (2023.01) [G06F 18/2155 (2023.01); G06F 18/24 (2023.01); G06N 3/08 (2013.01); G06V 10/75 (2022.01)] 18 Claims
OG exemplary drawing
 
1. A system comprising:
a hardware processor; and
a system memory storing a software code, a plurality of activation candidate functions for a last activation layer and a plurality of loss candidate functions for a loss layer;
the hardware processor configured to execute the software code to:
configure a neural network (NN) using a first combination including a first activation candidate function of the plurality of activation candidate functions for the last activation layer and a first loss candidate function of the plurality of loss candidate functions for the loss layer;
input a training dataset into the NN configured using the first combination;
receive, from the NN configured using the first combination, a first classification of the training dataset;
configure the NN using a second combination including a second activation candidate function of the plurality of activation candidate functions for the last activation layer and a second loss candidate function of the plurality of loss candidate functions for the loss layer;
input the training dataset, into the NN configured using the second combination;
receive, from the NN configured using the second combination, a second classification of the training dataset; and
determine, based on the first classification and the second classification, one of the first combination or the second combination as a preferred combination.