US 12,217,150 B2Granted

Data processing method and apparatus based on recurrent neural network, and device

AssigneeBeijing Baidu Netcom Science Technology Co., Ltd.

InventorsHuihuang Zheng, Xiang Lan, Yamei Li, Liujie Zhang, Fei Guo, Yanjun Ma, Dianhai Yu

Life of this patent

Estimating the term…

Loading patent…

Abstract

A data processing method and apparatus based on a recurrent neural network and a device are provided. The recurrent neural network includes multiple recurrent units, each recurrent unit includes multiple data processing nodes and a start node, at least one recurrent unit includes an end node, and at least one data processing node is included between the start node and the end node. During the processing of the first target processing object in a first recurrent unit, in a case that the first target processing object does not satisfy the first preset condition, the start node in the first recurrent unit is run to add a tag to the data processing nodes subsequent to the start node and stop addition of the tag in response to reaching the end node, and no processing is performed by the data processing nodes with the tag.

Claims17

Dependent claims

A data processing method based on a recurrent neural network, the recurrent neural network comprises a plurality of recurrent units, each recurrent unit comprises a plurality of data processing nodes, wherein each recurrent unit further comprises a first start node, at least one of the recurrent units further comprises a first end node, at least one of the data processing nodes is included between the first start node and the first end node, and the data processing method based on the recurrent neural network comprises:

inputting a plurality of target processing objects into the plurality of recurrent units, respectively;
sequentially processing the respective target processing objects by the plurality of data processing nodes in the respective recurrent units, to obtain a plurality of processing results output by the plurality of recurrent units;
wherein the plurality of recurrent units comprise a first recurrent unit, judging whether a first target processing object satisfies a first preset condition during the processing of the first target processing object in the first recurrent unit, and in a case that the first target processing object does not satisfy the first preset condition, no processing being performed by the data processing nodes between the first start node and the first end node during the processing of the first target processing object by the plurality of data processing nodes in the first recurrent unit;
wherein the method further comprises:
in a case that the first target processing object satisfies the first preset condition, no processing being performed by the first start node in the first recurrent unit and processing being performed on the first target processing object by the plurality of data processing nodes in the first recurrent unit;
wherein each recurrent unit further comprises a second start node, the at least one of the recurrent units further comprises a second end node, the second start node is located between the first start node and the first end node, and a second preset condition is provided with respect to each recurrent unit;
during the processing of the first target processing object in the first recurrent unit, the data processing method further comprises:
judging whether the first target processing object satisfies the second preset condition, in a case that the first target processing object satisfies the first preset condition;
no processing being performed by the data processing nodes between the second start node and the second end node during the processing of the first target processing object by the plurality of data processing nodes in the first recurrent unit, in a case that the first target processing object does not satisfy the second preset condition;
wherein the target processing objects comprise at least one of video files, audio files, or image files.

7 more elements

A data processing apparatus based on a recurrent neural network, the recurrent neural network comprising a plurality of recurrent units, each recurrent unit comprises a plurality of data processing nodes, wherein each recurrent unit further comprises a first start node, at least one of the recurrent units further comprises a first end node, at least one of the data processing nodes is included between the first start node and the first end node, and the data processing apparatus based on the recurrent neural network comprises:

a processor and a memory for storing one or more computer programs executable by the processor,
wherein when executing at least one of the computer programs, the processor is configured to perform operations comprising:
inputting a plurality of target processing objects into the plurality of recurrent units, respectively;
causing the plurality of data processing nodes in the respective recurrent units to sequentially process the respective target processing objects, to obtain a plurality of processing results output by the plurality of recurrent units;
wherein the plurality of recurrent units comprise a first recurrent unit,
when executing at least one of the computer programs, the processor is further configured to perform operations comprising:
judging whether a first target processing object satisfies a first preset condition during the processing of the first target processing object in the first recurrent unit, and in a case that the first target processing object does not satisfy the first preset condition, causing the data processing nodes between the first start node and the first end node to perform no processing during the processing of the first target processing object by the plurality of data processing nodes in the first recurrent unit;
wherein when executing at least one of the computer programs, the processor is further configured to perform operations comprising:
in a case that the first target processing object satisfies the first preset condition, no processing being performed by the first start node in the first recurrent unit and processing being performed on the first target processing object by the plurality of data processing nodes in the first recurrent unit;
wherein each recurrent unit further comprises a second start node, the at least one of the recurrent units further comprises a second end node, the second start node is located between the first start node and the first end node, and a second preset condition is provided with respect to each recurrent unit;
when executing at least one of the computer programs, the processor is further configured to perform operations comprising:
judging whether the first target processing object satisfies the second preset condition during the processing of the first target processing object in the first recurrent unit, in a case that the first target processing object satisfies the first preset condition;
causing the data processing nodes between the second start node and the second end node to perform no processing during the processing of the first target processing object by the plurality of data processing nodes in the first recurrent unit, in a case that the first target processing object does not satisfy the second preset condition;
wherein the target processing objects comprise at least one of video files, audio files, or image files.

11 more elements

A non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform a data processing method based on a recurrent neural network,

wherein the recurrent neural network comprises a plurality of recurrent units, each recurrent unit comprises a plurality of data processing nodes, wherein each recurrent unit further comprises a first start node, at least one of the recurrent units further comprises a first end node, at least one of the data processing nodes is included between the first start node and the first end node, and the computer is caused to perform the data processing method comprising:
inputting a plurality of target processing objects into the plurality of recurrent units, respectively;
sequentially processing the respective target processing objects by the plurality of data processing nodes in the respective recurrent units, to obtain a plurality of processing results output by the plurality of recurrent units;
wherein the plurality of recurrent units comprise a first recurrent unit, judging whether a first target processing object satisfies a first preset condition during the processing of the first target processing object in the first recurrent unit, and in a case that the first target processing object does not satisfy the first preset condition, no processing being performed by the data processing nodes between the first start node and the first end node during the processing of the first target processing object by the plurality of data processing nodes in the first recurrent unit;
wherein the computer is caused to perform the data processing method further comprising:
in a case that the first target processing object satisfies the first preset condition, no processing being performed by the first start node in the first recurrent unit and processing being performed on the first target processing object by the plurality of data processing nodes in the first recurrent unit;
wherein each recurrent unit further comprises a second start node, the at least one of the recurrent units further comprises a second end node, the second start node is located between the first start node and the first end node, and a second preset condition is provided with respect to each recurrent unit;
wherein the computer is caused to perform the data processing method further comprising:
judging whether the first target processing object satisfies the second preset condition during the processing of the first target processing object in the first recurrent unit, and in a case that the first target processing object satisfies the first preset condition;
and
causing the data processing nodes between the second start node and the second end node to perform no processing during the processing of the first target processing object by the plurality of data processing nodes in the first recurrent unit and in a case that the first target processing object does not satisfy the second preset condition;
wherein the target processing objects comprise at least one of video files, audio files, or image files.

9 more elements

Dependencies are inferred from the claim text.

Facts

Application: 17/248,131Filed 11 Jan 2021
Priority: CN 202010600160.528 Jun 2020
Published: US 2021/0406641 A130 Dec 2021
Granted: 4 Feb 2025
Term ends: Not estimatedThe source records don’t support an estimate.

Classification8 codes

G06N 3/044
Architecture, e.g. interconnection topology: recurrent networks, e.g. Hopfield networks
G06N 3/0442
Recurrent networks, e.g. Hopfield networks: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
G06N 3/045
Architecture, e.g. interconnection topology: combinations of networks
G06N 3/082
Learning methods: modifying the architecture, e.g. adding, deleting or silencing nodes or connections

Family

US 2021/0406641 A1Pre-grant publication · published 30 Dec 2021
US 12,217,150 B2This patent

From the US filing history. Foreign family members aren’t listed.

Timeline

Citations

Open elsewhere

Building an agent or pulling many patents? Use the API or MCP server instead of scraping pages. Get an API key