Attention LSTM-based reinforcement learning Agent knowledge reasoning method

A technology of reinforcement learning and knowledge reasoning, applied in the fields of reinforcement learning and deep learning, it can solve problems such as inapplicability of large-scale knowledge graphs and ineffective memory, and achieve the effect of realizing memory path screening, suppressing invalid states, and optimizing the reward mechanism.

Inactive Publication Date: 2020-12-22
SHANDONG ARTIFICIAL INTELLIGENCE INST
View PDF0 Cites 5 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

At present, the technology in this field involves knowledge reasoning based on first-order logic rules, which is only applicable to single-hop paths, knowledge reasoning based on random path ranking is not suitable for large-scale knowledge graphs, and knowledge reasoning based on FNN reinforcement learning Agent cannot effectively memorize reasoning paths

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Attention LSTM-based reinforcement learning Agent knowledge reasoning method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0032] The preprocessing operations in step a) include inference path information for statistical training, inference path information for statistical testing, and tokenization of entity relationships.

Embodiment 2

[0034] In step b), through the OpenKE-based TransH, TransE, TransR, DistinctMult, and CompLEx embedding models, the embedded word vector representation of the entity relationship in the triplet is obtained, and each entity and relationship is mapped to a dense continuous word vector.

Embodiment 3

[0036] In step c), through the formula P(S t+1 =s'|S t =s,max(A t )=a) Define the state transition equation at time t, where P is the probability of selecting a s' at time t+1, s' is the state variable at time t+1, and a is the state S at time t. t The maximum probability relation chosen, S t+1 is the state at time t+1, s is the entity associated with a, A t =P softmax (a|θ), θ is the network model parameter, through the formula Define the return function R(s t ), where e endThe final entity inferred for the relational path, e target target entity for relational path inference, e source is the set of entities in a given training path, e noanswer No node was found when reasoning the path, r + Indicates that the return value is a positive number, r - Indicates that the return value is negative.

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

An Attention LSTM-based reinforcement learning Agent knowledge reasoning method can effectively memorize a knowledge graph reasoning path through a bidirectional long-short-term memory network, and meanwhile, an attention mechanism weights the state of the memory path to obtain a state needing to be concerned, inhibits an invalid state, realizes memory path screening, and the problem that knowledge reasoning of a reinforcement learning Agent cannot effectively memorize a reasoning path is effectively solved. The LSTMAttention network model is used in the reinforcement learning Agent for relation path feature extraction, meanwhile, a return mechanism is optimized, and the reasoning precision of a knowledge graph reasoning algorithm based on reinforcement learning under multiple reference data sets is effectively improved.

Description

technical field [0001] The invention relates to the technical field of reinforcement learning and deep learning, in particular to an Attention-LSTM-based reinforcement learning agent knowledge reasoning method. Background technique [0002] Both the automatically built knowledge graph and the manually built graph face problems such as incompleteness, lack of knowledge, and judgment of the correctness of instances, making it difficult to apply to vertical search, question answering systems, and other fields. One of the solutions implements effective knowledge multi-hop reasoning on knowledge graphs to complete knowledge graphs, link predictions, and judge the correctness of instances. At present, the technology in this field involves that knowledge reasoning based on first-order logic rules is only applicable to single-hop paths, knowledge reasoning based on random path ranking is not suitable for large-scale knowledge graphs, and knowledge reasoning based on FNN reinforcemen...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06N3/04G06N5/04G06F16/36
CPCG06N3/049G06N5/04G06F16/367G06N3/044G06N3/08G06N3/006G06N5/02G06N7/01G06N3/045
Inventor 舒明雷刘浩王英龙刘辉陈超
Owner SHANDONG ARTIFICIAL INTELLIGENCE INST
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap