Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Machine learning algorithm-oriented prediction and training process unification method and device

A machine learning and algorithm technology, applied in the computer field, can solve difficult problems such as the unification of the algorithm training process and the prediction process

Inactive Publication Date: 2021-06-18
BEIJING REALAI TECH CO LTD
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] However, in order to achieve data privacy protection, it is difficult to unify the algorithm training process and prediction process in the longitudinal federated learning process of machine learning training algorithms

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
  • Machine learning algorithm-oriented prediction and training process unification method and device
  • Machine learning algorithm-oriented prediction and training process unification method and device
  • Machine learning algorithm-oriented prediction and training process unification method and device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0054] In order to facilitate the understanding of the unified method for the prediction and training process of the machine learning algorithm, this embodiment first introduces the machine learning algorithm.

[0055] The machine learning algorithm in this embodiment refers to a dynamic privacy protection machine learning algorithm, such as XGBoost (eXtreme Gradient Boosting, extreme gradient boosting) algorithm, GBDT (Gradient Boosting Decision Tree, gradient boosting decision tree) algorithm, LightGBM algorithm. Unless otherwise specified, the embodiments of the present disclosure use the XGBoost algorithm as an example of a dynamic privacy protection machine learning algorithm.

[0056] For the dynamic privacy-preserving machine learning algorithm, the object of the training process is a tensor structure, which can be easily indexed and easily generated. The object refers to the test_data and apply_data executed by different parties. The dynamic privacy protection machine ...

Embodiment 2

[0085] refer to image 3 , the embodiment of the present disclosure provides a unified device for the prediction and training process of machine learning algorithms, including:

[0086] The algorithm obtaining module 302 is used to obtain the first machine learning algorithm; wherein, the action object of the basic primitive in the first machine learning algorithm is a tree structure that cannot be accessed by indexing;

[0087] The primitive transformation module 304 is used for transforming the calculation flow of the basic primitive, so that the effect object of the basic primitive after transformation is a tree structure accessed in an index manner;

[0088] Algorithm conversion module 306, configured to convert the first machine learning algorithm into a second machine learning algorithm that meets the data privacy protection protocol based on the transformed basic primitives; wherein, the object of the second machine learning algorithm is accessed in an index manner A t...

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 relates to a machine learning algorithm-oriented prediction and training process unification method and apparatus. The method comprises the steps of obtaining a first machine learning algorithm, wherein the action object of the basic primitive in the first machine learning algorithm is a tree structure which cannot be accessed in an index mode; carrying out calculation flow transformation on the basic primitive, so that the action object of the transformed basic primitive is a tree structure accessed in an index mode; based on the transformed basic primitive, converting the first machine learning algorithm into a second machine learning algorithm satisfying a data privacy protection protocol, wherein the action object of the second machine learning algorithm is a tree structure accessed in an index mode, and the tree structure comprises internal nodes, leaf nodes and position indexes of the nodes; the second machine learning algorithm is used for the prediction task based on the tree structure. According to the invention, unification of a prediction process and a training process can be realized in a non-intrusive manner.

Description

technical field [0001] The present disclosure relates to the field of computer technology, and in particular to a unified method and device for predicting and training procedures of machine learning algorithms. Background technique [0002] Data privacy protection generally means that the characteristics or labels of the place where the data belongs (the party with permission) cannot be transmitted in plain text to the place where the data does not belong (the party without permission). To complete model training interactively, it is necessary to ensure data security and protect data privacy. The main technology for data privacy protection in current machine learning scenarios is federated learning. A common data privacy protection scenario is that feature data and label data are in two different parties (also known as vertical federated learning), and both parties need to complete machine learning without directly or indirectly leaking feature data / label data. Training al...

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): G06F21/62G06K9/62G06N20/00
CPCG06F21/6245G06N20/00G06F18/24323
Inventor 不公告发明人
Owner BEIJING REALAI TECH CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com