Patent · US Active

Apparatus and method for executing recurrent neural network and LSTM computations

US11531860B2 · kind B2 · utility

0Cited by

6References

28Claims

0Family size

Assignee

CAMBRICON (XI'AN) SEMICONDUCTOR CO., LTD. · CN

Inventors

Qi Guo · Shanghai, CN
Xunyu Chen · Zhuqiao, CN
Yunji Chen · Beijing, CN
Tianshi Chen · Beijing, CN

Key dates

Filing date	Oct 29, 2018
Grant date	Dec 20, 2022
Priority date	—
Expiry date	Jul 31, 2041

Classification

Technology area (CPC G)Physics
CPC primaryG06N3/084
WIPO fieldComputer technology
WIPO sectorElectrical engineering

Abstract

Aspects for Long Short-Term Memory (LSTM) blocks in a recurrent neural network (RNN) are described herein. As an example, the aspects may include one or more slave computation modules, an interconnection unit, and a master computation module collectively configured to calculate an activated input gate value, an activated forget gate value, a current cell status of the current computation period, an activated output gate value, and a forward pass result.

Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.