US 12,141,238 B2
Deep neural network hardener
Philip A. Sallee, South Riding, VA (US); and James Mullen, Leesburg, VA (US)
Assigned to Raytheon Company, Tewksbury, MA (US)
Filed by Raytheon Company, Tewksbury, MA (US)
Filed on Oct. 27, 2020, as Appl. No. 17/081,612.
Prior Publication US 2022/0129712 A1, Apr. 28, 2022
Int. Cl. G06N 3/063 (2023.01); G06F 18/22 (2023.01); G06F 18/2321 (2023.01); G06F 18/2415 (2023.01); G06N 3/04 (2023.01); G06N 3/08 (2023.01)
CPC G06F 18/24155 (2023.01) [G06F 18/22 (2023.01); G06F 18/2321 (2023.01); G06N 3/04 (2013.01); G06N 3/063 (2013.01); G06N 3/08 (2013.01)] 18 Claims
OG exemplary drawing
 
1. A computer-implemented method for content classification using a deep neural network (DNN) classifier, the method comprising:
obtaining content to be classified by the DNN classifier, the DNN classifier including a first classification layer; and
operating the DNN classifier to determine predicted class probabilities and corresponding confidences of the content using a second classification layer in place of the first classification layer, the second classification layer including a clustering classification layer that clusters based on a latent feature vector representation in latent feature space of the content, the confidences determined based on a distance between the latent feature vector representation and respective representative points of clusters in the latent feature space.