Tree recurrent neural network algorithm with attention mechanism batch training

Abstract

The invention discloses a tree recurrent neural network algorithm with attention mechanism batch training, and the algorithm comprises the steps: selecting sequence batch data of any length, and importing the data into a recurrent neural network (RNN); based on the RNN, establishing a long-short term memory network depending on gate vectors and memory units of all filial generation units, and obtaining a tree long-short term memory network; adding an attention mechanism based on the tree long-short term memory network to obtain an attention mechanism long-short term memory network; and carrying out batch training on sequence batch data through the attention mechanism long-short term memory network, and accelerating the training.

Description

technical field [0001] The invention relates to the field of deep learning in computer science, and in particular to the accelerated training of neural networks in deep learning. More specifically, the invention relates to a tree recurrent neural network algorithm with batch training of an attention mechanism. Background technique [0002] At this stage, for the virus monitoring of large-scale farms, there is no technical solution to use drone inspections to monitor the inside and outside of the farm for patrol virus sampling and monitoring; the isolation, collection and monitoring of farm entrances through non-contact collection still needs to be further resolved; how to Problems such as remote temperature measurement of breeding groups, how to conduct comprehensive non-contact collection and testing, and how to monitor breeding groups still need to be resolved; how to carry out intelligent control based on monitoring results, and timely, safe and thorough treatment to remov...

AI Technical Summary

Problems solved by technology

[0002] At this stage, for the virus monitoring of large-scale farms, there is no technical solution to use drone inspections to monitor the inside and outside of the farm for patrol virus sampling and monitoring; the isolation, collection and monitoring of farm entrances through non-contact collection still needs to be further resolved; how to Problems such as remote temperature measurement of breeding groups, how to conduct comprehensive non-contact collection and testing, and how to monitor breeding groups still need to be resolved; how to carry out intelligent control based on monitoring results, and timely, safe and thorough treatment to remove viruses still needs to be improved; therefore, It is necessary to propose a tree-recurrent neural network algorithm with batch training of attention mechanism to at least partially solve the problems existing in the prior art

Images