VAE-based medical care federated learning framework determination method

A deterministic method and federated technology, applied in the field of computer science, can solve problems such as poor effect of the global model, high communication cost of the federated learning framework, and slow convergence of the global model

Active Publication Date: 2021-12-14
UNIV OF ELECTRONICS SCI & TECH OF CHINA
View PDF5 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] Although federated learning is a very promising distributed learning framework, it still has many problems and challenges
The first problem is that the data distribution of each node under distributed training is not independent and identically distributed (non-IID), which will lead to two problems: the convergence of the global model will be very slow and the final global model effect will be Much less effective than centralized learning models
The second problem is the problem of data imbalance (imbalanced), which is particularly prominent in the field of healthcare, because abnormal data is much less likely to occur than normal data, such as in the data of fall detection, the data o

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
  • VAE-based medical care federated learning framework determination method
  • VAE-based medical care federated learning framework determination method
  • VAE-based medical care federated learning framework determination method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0058] This embodiment provides a determination method based on the VAE healthcare federated learning framework, which specifically includes the following steps:

[0059] Step 1: The medical institution obtains the initialized VAE model and initializes the anomaly detection model from the cloud server;

[0060] Step 2: Each medical institution trains and initializes the VAE model based on its own data set, and obtains the initial VAE model according to the selection of the data set fe Model and initialize VAE gen model; initialize the VAE fe The data set for model training includes all data, initialized VAE gen The dataset for model training includes sample data;

[0061] Step 3: The cloud server will obtain the initial VAE from each medical institution fe Model and initialize VAE gen The gradient of the model, after gradient aggregation; initialize the trained first generation VAE fe Model and first generation initialization VAE gen Models are distributed to all medica...

Embodiment 2

[0069] On the basis of Embodiment 1, the data set of the home gateway is acquired by the sensors in the wearable device, the sensors include angular velocity and acceleration sensors, and each sensor contains sequence information on the xyz three axes; The method of body data is to use Gramian Angular Field technology to convert it into two-dimensional image data.

[0070] Specifically, such as Figure 4 As shown, the left side is the original sequence data, and the right side is the 6-dimensional image data. The advantage of this processing method is that no matter how long the sequence data is, it can be unified into a 3D image of the same size, which is convenient for subsequent processing.

Embodiment 3

[0072] On the basis of Example 2, the VAE fe The function of the model is to extract low-dimensional features from the original high-dimensional data, and make the features obey the normal distribution; the VAE gen The function of the model is to generate more abnormal samples; the training process is as follows:

[0073] Step 41: The cloud server initializes the trained first generation VAE fe Model and first generation initialization VAE gen The model is distributed to all medical institutions participating in the training. The total number of medical institutions is k, and the data set owned by the kth medical institution is

[0074] Step 42: First Generation Initialize VAE fe The encoding end and decoding end of the model are denoted as E fe and D fe ; Initialize the VAE for the first generation gen The encoding end and decoding end of the model are denoted as E gen and D gen ;Each node needs to minimize the following two loss functions: this node refers to the m...

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 VAE-based medical care federated learning framework determination method, relates to the technical fields of computer science, machine learning and federated learning, and solves the three challenges in the federated learning-based medical care field: 1, the problems of final model performance reduction and slow convergence rates caused by non-IID data distribution; secondly, the problem of large model deviations caused by unbalanced distribution of positive and negative samples of medical data; and thirdly, the problem of difficult actual deployment caused by too large communication bandwidths consumed under federated learning. According to the invention, two lightweight VAEs are trained under a federated learning framework; then the trained VAEs are distributed to all nodes to be used for updating local data of the nodes, wherein updated local data has the characteristics that all the data is low in dimension, distribution is similar, and categories are balanced; and finally, an anomaly detection model is trained by using the data.

Description

technical field [0001] The present invention relates to the field of computer science, more specifically to the technical field of determination methods based on the VAE healthcare federated learning framework. Background technique [0002] In recent years, the advent of wearable technology has improved patients' lives and treatment experiences. Wearables and mobile devices are fundamentally changing the way we approach healthcare. Remote Patient Monitoring (RPM) is an implementation of the Internet of Medical Things (IoMT) that helps in delivering high-quality care and timely remote intervention to avert health crises. In order to obtain a high-quality remote anomaly detection system (such as abnormal heart rate detection, fall detection, etc.), it must be learned through a large amount of data. However, data in the medical field is very sensitive and private, which means that data between medical institutions cannot be shared. Therefore, unlike the traditional method of ...

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): G06N3/04G06N3/08G06N20/00G16H50/70
CPCG06N3/08G06N20/00G16H50/70G06N3/047
Inventor 杨浩淼葛孟雨金禹樵张益李佳晟王宇卢锐恒汤殿华李宇博李发根
Owner UNIV OF ELECTRONICS SCI & TECH OF CHINA
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