公开(公告)号：US20200167655A1

公开(公告)日：2020-05-28

申请号：US16697646

申请日：2019-11-27

Applicant: Electronics and Telecommunications Research Institute

Inventor： Jun Yong PARK

IPC: G06N3/08 ,  G06N3/04 ,  G06N3/10

Abstract: Disclosed are a method and apparatus for generating an ultra-light binary neural network which may be used by an edge device, such as a mobile terminal. A method of re-configuring a neural network includes obtaining a neural network model on which training for inference has been completed, generating a neural network model having a structure identical with the neural network model on which the training has been completed, performing sequential binarization on an input layer and filter of the generated neural network model for each layer, and storing the binarized neural network model. The method may further include providing the binarized neural network model to a mobile terminal.

公开(公告)号：US20200167659A1

公开(公告)日：2020-05-28

申请号：US16696061

申请日：2019-11-26

Applicant: Electronics and Telecommunications Research Institute

Inventor： Yong Hyuk MOON ,  Jun Yong PARK ,  Yong Ju LEE

IPC: G06N3/08 ,  G06N5/04 ,  G06N20/00 ,  G06F17/16

Abstract: Provided are a device and method for training a neural network. The method includes generating a candidate solution set by modifying a candidate solution which represents a basic neural network model in a variable-length string form, acquiring first candidate solutions by performing architecture variation-based unsupervised learning with a plurality of candidate solutions selected from the candidate solution set, selecting a neural network model represented by a first candidate solution which satisfies targeted effective performance as a first neural network model, acquiring second candidate solutions by performing selective error propagation-based supervised learning with the first neural network model, and selecting a neural network model represented by a second candidate solution which satisfies the targeted effective performance as a final neural network model.

公开(公告)号：US20210125063A1

公开(公告)日：2021-04-29

申请号：US17038894

申请日：2020-09-30

Applicant: Electronics and Telecommunications Research Institute

Inventor： Jun Yong PARK

IPC: G06N3/08 ,  G06N3/04

Abstract: A method for generating a binary neural network may comprise extracting real-value filter weights from a first neural network for which inference training has been completed; performing a binary orthogonal transform on the filter weights; and generating a second neural network using binary weights calculated according to the binary orthogonal transform.