MambaADv2: Evolving Duality-enhanced State Space Model for Unsupervised Anomaly Detection

2026-06-22

AI Summary

Existing convolutional neural networks struggle to model long-range dependencies, while Transformers suffer from high computational complexity, both limiting performance in unsupervised anomaly detection. To address these challenges, this work proposes MambaADv2, a novel framework that integrates a pretrained encoder with a Mamba-inspired decoder. The core innovation lies in a multi-scale Duality-enhanced State Space (DSS) module, which combines parallel cascaded HSS blocks with frequency-domain enhanced convolutions and employs a semantic-adaptive progressive scanning strategy. By leveraging position-aware state space modeling, linear recurrence, and dual-path parallel computation, the method effectively captures both global dependencies and local details. This enables highly accurate and efficient unsupervised anomaly detection across multiple categories, achieving linear computational complexity while excelling in long-range modeling capability.

Abstract

While recent advancements in anomaly detection have demonstrated the efficacy of CNN- and Transformer-based approaches, these architectures face inherent limitations: CNNs struggle to capture long-range dependencies, whereas Transformers suffer from quadratic computational complexity. Consequently, Mamba-based architectures have attracted considerable attention, as they successfully combine superior long-range dependency modeling with linear computational complexity. By critically rethinking the structural evolution across the Mamba lineage 1-3 series, this paper proposes MambaADv2, a framework tailored for multi-class unsupervised anomaly detection. MambaADv2 comprises a pre-trained encoder and a Mamba-inspired decoder, equipped with Duality-enhanced State Space (DSS) modules across multiple scales. The proposed DSS module effectively models both global dependencies and local representations by integrating parallel-cascaded Hybrid State Space (HSS) blocks and frequency-enhanced convolution operations. The structure of the Hybrid State Space (HSS) block is tailored by following the SSD-based Mamba lineage and incorporating Mamba3-style position-aware state-space modeling, leveraging the dual computational paths of linear recurrence and parallel matrix formulation to model local continuity and global contextual comparison, thereby better serving the core anomaly detection objective of precisely reconstructing normal representations while magnifying anomalous deviations. Additionally, we propose a semantics-adaptive progressive scanning strategy that decays scanning complexity along the feature pyramid.

Problem

• anomaly detection
• long-range dependencies
• computational complexity
• unsupervised learning
• multi-class

Innovation

• Mamba
• State Space Model
• Unsupervised Anomaly Detection
• Duality-enhanced
• Hybrid State Space