Sensitive long-term and short-term memory method based on output variation differentiation

Abstract

The invention discloses a sensitive long-term and short-term memory method based on output variation differentiation. The objective of the invention is to improve the response capability of a traditional LSTM neural network to short-time information. According to the method, the neural unit of the long-term and short-term memory network with increased information sensitivity is added, so that theresponse capability of the network to short-term information can be well improved. The application real-time performance of the network is improved. More perfect real-time analysis can be carried out.Micro-actions and other contents can be further analyzed. The application value is improved.

Description

technical field

[0001] The invention relates to the field of long-short-term memory network, in particular to a sensitive long-short-term memory method based on output variation differentiation. Background technique

[0002] Artificial intelligence is one of the three important disciplines in the 21st century and an important support for national science, economy, and people's livelihood. Among them, the long-short-term memory network (LSTM) is an important algorithm for memory-based recognition, which has been recognized in many aspects including semantics, actions, texts, etc., and has very good value.

[0003] The existing long-term short-term memory network still has a major problem, that is, it uses long-term short-term memory to improve the analysis ability of information in the long-term sequence of the entire video, but there is no response to short-term information at all. ability, which makes the existing long-short-term memory network can only be used for post-ev...

Images