Deep neural network process interpretation device based on hierarchical prediction evolution

The invention discloses a deep neural network process interpretation device based on hierarchical prediction evolution. The deep neural network process interpretation device comprises an interpreter; the interpreter comprises an extraction layer, a reconstruction layer and a quantization module; the...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: YU ZHENYAO, YU ZIYI, LIU ZITU, LI JIAWANG, LIU YUE
Format: Patent
Sprache:chi ; eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The invention discloses a deep neural network process interpretation device based on hierarchical prediction evolution. The deep neural network process interpretation device comprises an interpreter; the interpreter comprises an extraction layer, a reconstruction layer and a quantization module; the extraction layer is used for acquiring prediction data of the deep neural network hidden layer and performing mapping processing on the prediction data to obtain probability distribution data; the reconstruction layer is used for processing and calculating the probability distribution data to obtain reconstruction prediction data, and obtaining fidelity data of the probability distribution data based on the reconstruction prediction data; and the quantization module is used for carrying out quantitative sorting processing on the probability distribution data to obtain sorting matrix data of probability distribution, and obtaining confidence data of prediction data based on the sorting matrix data to realize predic