ELECTRONIC DEVICE AND CONTROL METHOD THEREOF
Disclosed is an electronic device. The present electronic device includes: a memory; and a processor which quantizes a neural network, trained on the basis of deep learning, to generate a quantized neural network, and stores the quantized neural network in the memory, wherein the processor quantizes...
Gespeichert in:
Hauptverfasser: | , , , |
---|---|
Format: | Patent |
Sprache: | eng ; fre ; kor |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | Disclosed is an electronic device. The present electronic device includes: a memory; and a processor which quantizes a neural network, trained on the basis of deep learning, to generate a quantized neural network, and stores the quantized neural network in the memory, wherein the processor quantizes, in preset first bit units, trained connection strengths between neurons of the trained neural network, inverse-quantizes the quantized connection strengths in preset second bit units, retrains the inverse-quantized connection strengths, and quantizes the retrained connection strengths in the preset first bit units.
La présente invention concerne un dispositif électronique. Le dispositif électronique de la présente invention comprend : une mémoire ; et un processeur qui quantifie un réseau neuronal, formé sur la base d'un apprentissage profond, pour générer un réseau neuronal quantifié, et stocke le réseau neuronal quantifié dans la mémoire, le processeur quantifiant, dans des premières unités de bits prédéfinies, des forces de connexion formées entre des neurones du réseau neuronal formé, quantifie en inverse les forces de connexion quantifiées dans des secondes unités de bits prédéfinies, reforme les forces de connexion à quantification inverse, et quantifie les forces de connexion reformées dans les premières unités de bits prédéfinies.
전자 장치가 개시된다. 본 전자 장치는 메모리 및 딥 러닝에 기초하여 학습된 뉴럴 네트워크(neural network)를 양자화하여 양자화된 뉴럴 네트워크를 생성하고, 양자화된 뉴럴 네트워크를 메모리에 저장하는 프로세서를 포함하고, 프로세서는 학습된 뉴럴 네트워크의 뉴런(neuron)들 사이의 학습된 연결 강도(trained connection strength)를 기설정된 제1 비트 단위로 양자화하고, 양자화 된 연결 강도를 기설정된 제2 비트 단위로 역 양자화하며, 역 양자화된 연결 강도를 재학습(retraining)하고, 재학습된 연결 강도를 기설정된 제1 비트 단위로 양자화한다. |
---|