Unlock instant, AI-driven research and patent intelligence for your innovation.

Reinforcement learning framework based on P2P network

A P2P network and reinforcement learning technology, applied in neural learning methods, biological neural network models, neural architectures, etc., can solve the problems of low threshold and high upper limit, and achieve the effect of high use limit and flexible application

Pending Publication Date: 2022-06-28
SHANGHAI MATRIXELEMENTS TECH CO LTD
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] The technical problem to be solved by the present invention is to provide a P2P network-based reinforcement learning that is flexible and compatible with various large reinforcement learning algorithms, with a low threshold, everyone can participate, a high upper limit, and can solve many big problems. frame

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
  • Reinforcement learning framework based on P2P network
  • Reinforcement learning framework based on P2P network
  • Reinforcement learning framework based on P2P network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0024] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] This embodiment provides a reinforcement learning framework based on a P2P network, such as figure 1 As shown, the reinforcement learning framework based on the P2P network is provided by MetisNet with the ability to initiate computing tasks on P2P. On top of this, a reinforcement learning framework called Metis0 is constructed, and other computing frameworks can also be built based on MetisNet, such as privacy computing framework; that is, the reinforcement learning framework is constructed on the P2P network provided by the MetisNet platform, and the MetisNet platform provides a program for discovering computing nodes and performing dynamic networking on the computing nodes; specifically, the MetisNet platform discovers by publishing tasks Compute nodes and perform dynamic networking, and select the trainer module according to the runt...

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

The invention discloses a reinforcement learning framework based on a P2P network. The reinforcement learning framework comprises a simulator module, a trainer module and an evaluator module, the simulator modules exist in each computing node, a current optimal neural network model is loaded, respective self-play training is performed to generate a large amount of data, and the generated data is uploaded to a public data node; the trainer module samples data from the public data node, trains a new neural network model, and stores the trained neural network model in the Model Pool; and the evaluator module loads the new neural network model from the Model Pool, so that the new neural network model and the current optimal neural network model are subjected to multiple battles, the battle rate is counted, and the new neural network model with the high battle rate is iterated into the current optimal neural network model. The method is low in threshold, can be participated by everyone, is high in upper limit, can solve many big problems, is flexible, and is compatible with various big reinforcement learning algorithms.

Description

technical field [0001] The invention relates to the field of computer software, in particular to a reinforcement learning framework based on a P2P network. Background technique [0002] Since the great success of DeepMind's work on AlphaGo, AlphaZero, etc., it has been realized that this computing paradigm can solve a large number of problems. Many teams have released their own deep reinforcement learning libraries or frameworks, but some of these libraries are suitable for training on a single machine (multi-card), and some are suitable for distributed training on a local area network. [0003] There is a situation: people get together in a circle for reasons such as interest and want to accomplish one thing together (such as training an agent that can provide chess decisions), they have their own ordinary machines, but there is no large-scale cluster, and they are not in the same network . How to accomplish this task? [0004] The existing reinforcement learning framewo...

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
IPC IPC(8): G06N3/04G06N3/08G06N5/04H04L67/104
CPCG06N3/04G06N3/08G06N5/046H04L67/104
Inventor 刘辉
Owner SHANGHAI MATRIXELEMENTS TECH CO LTD