Multi-scale adaptive gating mamba plus network construction method and device

By using a multi-scale adaptive gated MambaPlus network, the problems of insufficient multi-scale information utilization and poor feature fusion adaptation in Mamba-like models are solved, achieving efficient feature representation and robustness improvement, and making it suitable for complex scenarios such as time series analysis and fault diagnosis.

CN122087742BActive Publication Date: 2026-07-10UNIV OF JINAN
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
UNIV OF JINAN
Filing Date
2026-04-27
Publication Date
2026-07-10

AI Technical Summary

Technical Problem

Existing Mamba-like models suffer from insufficient utilization of multi-scale information, poor adaptive feature fusion, weak module synergy, low robustness in complex scenarios, and lack a unified network structure with explicit multi-scale feature extraction, adaptive gating fusion, and feature enhancement mechanisms.

Method used

A multi-scale adaptive gated MambaPlus network is constructed. Through multi-scale feature extraction, hierarchical adaptive gating fusion, and cross-scale attention enhancement, combined with enhanced normalization, training-time noise injection, AdamW optimizer, and cosine annealing learning rate scheduling, a closed-loop collaborative architecture of multi-scale feature extraction, hierarchical gating fusion, and cross-scale attention enhancement is formed.

Benefits of technology

It significantly improves the model's ability to represent multi-scale features of complex patterns, enhances the model's adaptability and robustness, improves recognition accuracy and generalization performance, and reduces model design and deployment costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122087742B_ABST
    Figure CN122087742B_ABST
Patent Text Reader

Abstract

The application discloses a multi-scale adaptive gating MambaPlus network construction method and device, belongs to the technical field of artificial intelligence and machine learning, and comprises the following steps: initializing network configuration and constructing a basic structure; pre-processing input data to generate a standard data set; mapping to a hidden space through an input mapping layer, extracting backbone features from a Mamba main body; constructing at least two parallel scale branches in the hidden space to obtain a multi-scale feature set; inputting the backbone features and the multi-scale features into an adaptive gating module, dynamically distributing weights and adaptively fusing through a hierarchical gating mechanism to generate fused features; then, cross-scale attention enhancement and feedforward enhancement are performed, and residual superposition is performed to generate final discriminative features; and finally, class prediction and model training evaluation are completed. While retaining the advantages of Mamba long-range dependence modeling, the application solves the problems of insufficient multi-scale information utilization, poor feature fusion adaptability and low robustness in complex scenes.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • A multidirectional feature fusion network-based system for efficient object detection

    AU2021102692A4

  • Multivariate time-series long-term forecasting based on multi-scale temporal feature enhancements

    US20250390715A1