Joint learning framework based on cooperation of cloud server and IoT equipment
A cloud server and equipment technology, applied in the field of neural network and federated learning, can solve the problems of poor neural network performance, long average execution time, low prediction accuracy, etc., achieve good performance, improve prediction accuracy and average inference execution time, the effect of protecting data privacy
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Publication Date
- 2020-09-04
Smart Images

Figure 1 
Figure 2 
Figure 3
Abstract
Description
technical field
[0001] The invention relates to the technical field of neural network and federated learning, in particular to a BranchyNet-based cloud server and terminal IoT device group collaborative reasoning and joint learning framework. Background technique
[0002] With the development of Internet of Things technology, the arrival of 5G communication and the increasing popularity of various IoT devices, the massive data generated by the IoT terminal in the future will soon exceed the data volume of the existing Internet. Edge and terminal intelligence are attracting more and more attention, and artificial intelligence technology can be effectively used to process data generated by massive devices. In recent years, the computing power of IoT devices has been continuously improved. In particular, AI chips dedicated to neural network computing have gradually become standard equipment for mid-to-high-end devices, and artificial intelligence has gradually migrated from the...
Examples
Embodiment 1
[0039] See attached Figure 4 , the present invention adopts the joint learning framework of cloud server and IoT device collaboration of distributed AI system architecture, and its joint learning specifically includes the following steps:
[0040] Step 1: Design the neural network model
[0041] Design a suitable BranchyNet neural network model according to the computing and storage resources of cloud servers and terminal IoT (Internet of Things, Internet of Things) devices. The Branch part of the cloud, the Trunk and Branch part.
[0042] Step 2: Cloud offline training
[0043] The BranchyNet model is trained on the cloud with public datasets, with the purpose of pre-training to obtain the weight of the initial BranchyNet neural network model.
[0044] The training method of the present invention is described below by taking cross entropy (cross entropy) as the loss function of the neural network as an example. The cloud offline training in the first stage is a joint opt...