An Iterative Random Pruning Method and System Based on Meta-Transfer Learning
By employing an iterative random pruning method based on meta-transfer learning, and utilizing pre-training on large-scale datasets and meta-training on small-sample datasets, combined with iterative random pruning and retraining, the problems of overfitting and large parameter count in deep learning models under small-sample scenarios are solved, achieving efficient model compression and improved accuracy.
CN117787377BActive Publication Date: 2026-09-01HANGZHOU DIANZI UNIV
2 Cites 0 Cited by
Patent Information
- Application Number
- CN202310610163.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-05-24
- Publication Date
- 2026-09-01
- Estimated Expiration
- 2043-05-24
AI Technical Summary
Technical Problem
Existing deep learning network models do not perform well with pruning methods in small sample scenarios, are difficult to deploy effectively on embedded mobile devices, and suffer from overfitting issues.
Method used
An iterative random pruning method based on meta-transfer learning is adopted. The network is pre-trained on a large-scale dataset, and meta-training and iterative random pruning are performed using a small sample dataset to reduce the number of model parameters. Accuracy is restored through retraining.
Benefits of technology
To improve model accuracy in small sample scenarios, reduce the number of parameters, alleviate overfitting, compress the model, and enhance its adaptability to small sample datasets.
✦ Generated by Eureka AI based on patent content.
Abstract
This invention belongs to the field of neural network technology, specifically relating to an iterative random pruning method and system based on meta-transfer learning. The pruning method of this invention includes the following steps: S1, a transfer learning step, pre-training the convolutional neural network using a large-scale dataset to initialize the network parameters; S2, a meta-learning step, meta-training the network using a small sample dataset; S3, a pruning step, pruning the network weights after meta-training; S4, a testing step, testing the network using a small sample dataset; and S5, a fine-tuning stage, retraining the network to restore its accuracy. Compared to existing technologies, this invention can reduce the number of parameters in a convolutional neural network with small sample datasets while maintaining the accuracy of the network model, achieving the goal of model compression.
Need to check novelty before this filing date? Find Prior Art