Deep learning data sharing method based on block chain mode, and storage medium

Abstract

The invention provides a deep learning data sharing method based on a block chain mode, and a storage medium. The method comprises the steps of storing deep learning data and a target modeling model in all nodes of a block chain network, wherein the deep learning data is shared in the block chain network, and the target modeling model is protected; splitting the deep learning data into metadata with corresponding data attribute tags; when a deep learning request corresponding to the target modeling model is received, determining the corresponding data attribute tags according to the target modeling model, and obtaining the corresponding metadata; and according to the corresponding metadata, extracting the deep learning data corresponding to the target modeling model to perform learning. The method has the advantages that a deep learning mode is optimized, and one step of reinforcement learning can be achieved, so that the learning efficiency can be remarkably improved; and furthermore,the learning data can be reused, and the learning cost and time can be saved.

Description

technical field [0001] The invention relates to the field of deep learning data, in particular to a deep learning data sharing method and a storage medium based on a block chain. Background technique [0002] The current common machine deep learning solution is the general modeling algorithm used by the Google DeepMind team. Using the deep learning model of convolutional neural network, several robots are connected to each other and share the experience and trial data of mutual learning. After nearly a million actions, the networked machines will gradually start to correct themselves and achieve the goal of self-learning. Effect. Then these data will be stored as a single experimental data of this learning method, and cannot be used as supporting data for other behavioral learning, that is, data reuse cannot be realized. [0003] At the same time, since the machine is learning independently, a large amount of data will be retained for each learning, and these data are uniq...

AI Technical Summary

Problems solved by technology

[0004] Furthermore, in the existing machine deep learning scheme, each learning needs to start from scratch, even if many parts of it have been mastered, it needs to be relearned, and it is impossible to achieve targeted direct learning of the sample data corresponding to the target model, resulting in inaccurate learning. Necessary waste of time and cost

Images