State-space Model Research Articles

ModelS4Apnea: leveraging structured state space models for efficient sleep apnea detection from ECG signals

ModelS4Apnea: leveraging structured state space models for efficient sleep apnea detection from ECG signals

DSMT-Net: Dual-student mean teacher network with pixel-level pseudo-label optimization for semi-supervised medical image segmentation.

DSMT-Net: Dual-student mean teacher network with pixel-level pseudo-label optimization for semi-supervised medical image segmentation.

Optimizing control strategies for DC-DC boost converters: Real-time application of an adaptive gain scheduled ISA-PI controller with hybrid state-space and linear parameter-varying modelling

This paper introduces an innovative sophisticated control scheme for a DC-DC boost converter (DCBC), employing an adaptive gain scheduled ISA-PI controller. Addressing the inherent non-minimum phase behaviour arising from a right-half plane zero and the complexities associated with nonlinear dynamics during continuous conduction mode (CCM), the proposed adaptive gain scheduled ISA-PI controller incorporates the distinct adjustable parameter within the controller structure. This parameter is instrumental in enhancing the adaptability of the controller to varied operating conditions. The adaptive ISA-PI controller seamlessly integrates the real-time duty cycle value, replacing traditional tuning variables with precision. The dynamic adjustment of this sole controllable parameter is facilitated through a carefully designed look-up table, employing the loop-shaping method. Verification of the proposed control system’s effectiveness is conducted using MATLAB/Simulink, incorporating a comprehensive comparative analysis against single proportional integral (PI) controllers. The assessment centres on evaluating the system’s precision in tracking desired signals and regulating plant process variables with optimal efficiency, minimizing delays and overshoot. Experimental validation is further undertaken using MATLAB/Simulink/Stateflow on a dSPACE Real-time-interface (RTI) 1007 processor, DS2004 High-Speed A/D, and CP4002 Timing and Digital I/O boards. The experimental results confirm the superior performance of the proposed adaptive gain schedule ISA-PI controller, which has a unique configurable parameter. This controller demonstrated a twofold improvement in tracking speed and significantly improved disturbance rejection, confirming its effectiveness.

Optimal model-based insulin bolus advisor for subjects with type 1 diabetes using continuous glucose monitoring.

Optimal model-based insulin bolus advisor for subjects with type 1 diabetes using continuous glucose monitoring.

Modeling Hysteretically Nonlinear Piezoelectric Composite Beams

This paper presents a modeling framework for hysteretically nonlinear piezoelectric composite beams using functional differential equations (FDEs). While linear piezoelectric models are well established, they fail to capture the complex nonlinear behaviors that emerge at higher electric field strengths, particularly history-dependent hysteresis effects. This paper develops a cascade model that integrates a high-dimensional linear piezoelectric composite beam representation with a nonlinear Krasnosel’skii–Pokrovskii (KP) hysteresis operator. The resulting system is formulated using a state-space model where the input voltage undergoes a history-dependent transformation. Through modal expansion and discretization of the Preisach plane, we derive a tractable numerical implementation that preserves essential nonlinear phenomena. Numerical investigations demonstrate how system parameters, including the input voltage amplitude, and hysteresis parameters significantly influence the dynamic response, particularly the shape and amplitude of limit cycles. The results reveal that while the model accurately captures memory-dependent nonlinearities, it depends on numerous real and distributed parameters, highlighting the need for efficient reduced-order modeling approaches. This work provides a foundation for understanding and predicting the complex behavior of piezoelectric systems with hysteresis, with potential applications in vibration control, energy harvesting, and precision actuation.

Prescribed Performance Load Frequency Control for Regional Interconnected Power System Under Energy Storage System Output Constraints

This study addresses the issue of frequency instability caused by an imbalance between load power and generation power in a power system. A state-space model of a two-area power system including a thermal power plant is first established, incorporating the output power limitations of the energy storage system, which is the actuator for frequency control. Under input saturation constraints, a frequency control strategy based on a prescribed performance control technique is proposed. This strategy not only ensures frequency stability but also achieves an optimal transient response curve. The proposed control strategy is theoretically validated and numerically simulated, demonstrating its effectiveness in suppressing frequency variations in power systems under constraints regarding the output power of the energy storage system.

A Hybrid MIL Approach Leveraging Convolution and State-Space Model for Whole-Slide Image Cancer Subtyping

Precise identification of cancer subtypes from whole slide images (WSIs) is pivotal in tailoring patient-specific therapies. Under the weakly supervised multiple instance learning (MIL) paradigm, existing techniques frequently fall short in simultaneously capturing local tissue textures and long-range contextual relationships. To address these challenges, we introduce ConvMixerSSM, a hybrid model that integrates a ConvMixer block for local spatial representation, a state space model (SSM) block for capturing long-range dependencies, and a feature-gated block to enhance informative feature selection. The model was evaluated on the TCGA-NSCLC dataset and the CAMELYON16 dataset for cancer subtyping tasks. Extensive experiments, including comparisons with state-of-the-art MIL methods and ablation studies, were conducted to assess the contribution of each component. ConvMixerSSM achieved an AUC of 97.83%, an ACC of 91.82%, and an F1 score of 91.18%, outperforming existing MIL baselines on the TCGA-NSCLC dataset. The ablation study revealed that each block contributed positively to performance, with the full model showing the most balanced and superior results. Moreover, our visualization results further confirm that ConvMixerSSM can effectively identify tumor regions within WSIs, providing model interpretability and clinical relevance. These findings suggest that ConvMixerSSM has strong potential for advancing computational pathology applications in clinical decision-making.

CT-Mamba: A hybrid convolutional State Space Model for low-dose CT denoising.

CT-Mamba: A hybrid convolutional State Space Model for low-dose CT denoising.

Startup Process of Pumped Storage Unit for Avoiding S-Shaped Region Based on Geometric Perspective Method

This paper aims to study the mechanism of avoiding the S-shaped region (S-shaped region, SFR) during the startup of pumped storage units (pumped storage units, PSUs). Firstly, the state space model of the PSU in frequency mode is built using the transfer coefficient of the pump turbine. Then, according to the characteristics of the SFR, the accurate range of the SFR is determined in the full characteristic curve. Finally, combined with a specific power station, this paper proposes a novel geometric perspective method to reveal the underlying mechanism for avoiding the SFR during the startup of PSUs. The core innovation lies in establishing, for the first time, the precise spatial relationship (positioning and distance) between the no-load operating point and the upper boundary of the SFR, thereby identifying two critical necessary and sufficient conditions for successful startup avoiding instability. Based on this mechanism, the critical state of PSUs entering the SFR and the influence of operation points on the startup stability that the PSU is putting into PID control are analyzed using the Hopf bifurcation principle. The results show that two conditions need to be met when the PSU starts up to avoid the SFR. One is that the system operation point is in the stable region, and the other is that the speed overshoot is less than the critical speed overshoot. The speed overshoot is the direct cause of the unit entering the SFR, leading to startup failure. When the PSU is started up and put into proportional–integral–derivative (proportional–integral–derivative, PID) control, a certain margin of flow and guide vane opening will help reduce the speed overshoot and prevent the unit from entering the SFR.

SSM-based detection of rice seedling deficiency

In large-scale rice cultivation, seedling deficiency is a common issue that significantly impacts timely replanting decisions. Traditional manual inspection methods are inefficient and labor-intensive, highlighting the need for an automated and accurate detection approach. This study proposes a rice seedling deficiency detection method based on a state space model, aiming to improve detection precision for small seedlings. With a dual-branch feature extraction module built upon the State Space Model (Mamba), and the wavelet convolution transform, enhances the detection accuracy on the self-constructed rice seedling deficiency dataset. Experiments show that the proposed optimized model achieves a mAP50 of 78%, outperforming other baseline models. The results indicate the effectiveness and practicality of the approach, offering a novel and efficient solution for detecting missing seedlings in rice fields.

Few-shot network intrusion detection method based on multi-domain fusion and cross-attention.

Deep learning methods have achieved remarkable progress in network intrusion detection. However, their performance often deteriorates significantly in real-world scenarios characterized by limited attack samples and substantial domain shifts. To address this challenge, we propose a novel few-shot intrusion detection method that integrates multi-domain feature fusion with a bidirectional cross-attention mechanism. Specifically, the method adopts a dual-branch feature extractor to jointly capture spatial and frequency domain characteristics of network traffic. The frequency domain features are obtained via two-dimensional discrete cosine transform (2D-DCT), which helps to highlight the spectral structure and improve feature discriminability. To bridge the semantic gap between support and query samples under few-shot conditions, we design a dual-domain bidirectional cross-attention module that enables deep, task-specific alignment across spatial and frequency domains. Additionally, we introduce a hierarchical feature encoding module based on a modified Mamba architecture, which leverages state space modeling to capture long-range dependencies and temporal patterns in traffic sequences. Extensive experiments on two benchmark datasets, CICIDS2017 and CICIDS2018, demonstrate that the proposed method achieves accuracy of 99.03% and 98.64% under the 10-shot setting, outperforming state-of-the-art methods. Moreover, the method exhibits strong cross-domain generalization, achieving over 95.13% accuracy in cross-domain scenarios, thereby proving its robustness and practical applicability in real-world, dynamic network environments.

DIPathMamba: A domain-incremental weakly supervised state space model for pathology image segmentation.

DIPathMamba: A domain-incremental weakly supervised state space model for pathology image segmentation.

DehazeMamba: large multi-modal model guided single image dehazing via mamba

Deep neural networks have achieved significant success in image dehazing. However, existing backbones face an irreconcilable trade-off between the global receptive field and computational efficiency, hindering further applications. State space models, such as Mamba, offer a potential solution to this conflict by modeling long-range dependencies with linear complexity. Although Mamba is well-suited for sequential tasks (e.g., natural language processing), it still encounters challenges when applied to low-level vision tasks. In this work, we propose a large multi-modal model (LMM) guided, Mamba-based image dehazing method (DehazeMamba). It enhances the standard Mamba architecture by incorporating image quality priors provided by the LMM and a channel attention mechanism. Additionally, we present a challenging image dehazing dataset and conduct new benchmark studies based on the LMM, evaluating hazy images and dehazing results by simulating human perception. Our experimental results demonstrate that our dataset exhibits superior haze quality, and our method outperforms current state-of-the-art (SOTA) dehazing methods by achieving a performance improvement of more than 5% on both the O-Haze and Dense-Haze datasets.

Medical image translation with deep learning: Advances, datasets and perspectives.

Medical image translation with deep learning: Advances, datasets and perspectives.

Two-stage Mamba-based diffusion model for image restoration

Image restoration is fundamental in computer vision to restore high-quality images from degraded ones. Recently, models such as the transformer and diffusion have shown notable success in addressing this challenge. However, transformer-based methods face high computational costs due to quadratic complexity, while diffusion-based methods often struggle with suboptimal results due to inaccurate noise estimation. This study proposes Diff-Mamba, a two-stage adaptive Mamba-based diffusion model for image restoration. Diff-Mamba integrates the linear complexity state space model (SSM, also known as Mamba) into image restoration, expanding its applicability to visual data generation. Diff-Mamba mainly consists of two parts: the diffusion state space model (DSSM) and the diffusion feedforward neural network (DFNN). DSSM combines Mamba’s high efficiency with the representative power of diffusion models, enhancing both inference and training. DFNN regulates the information flow, enabling each depthwise convolutional layer to focus on the details of image, thus learning more effective local structures for image restoration. The study’s findings, verified through extensive experiments, indicate that Diff-Mamba outperforms both diffusion-based and transformer-based methods in image deraining, denoising, and deblurring, demonstrating competitive restoration performance with various commonly used datasets. Code is available at https://github.com/maluan-ml/Diff-Mamba.

Diagnosis of Alzheimer’s disease using brain -FDG PET imaging based on a state space model

In recent studies on Alzheimer’s disease (AD), various network models have shown significant potential in disease prediction. However, traditional CNNs often rely on parameterized loss functions, limiting the robustness of these models. Additionally, their high computational complexity increases resource demands. To address these challenges, this study proposes a novel prediction model that integrates the strengths of ViTs and the MedMamba module. First, this study draws on the SS-Conv-SSM module from the MedMamba model, which processes image branches in parallel to extract richer and more refined features. Building on this, we optimized the original purely convolutional structure into a hybrid architecture combining convolution and Transformer layers. This not only reduces the computational burden and enhances operational efficiency but also improves the model’s ability to capture global features. Moreover, we introduced a new self-attention mechanism into the model’s MDTA module, reducing the computational complexity from quadratic to linear. This allows the model to maintain high performance while achieving more lightweight and efficient operations. The final experimental results demonstrate that this model outperforms current state-of-the-art methods in predicting Alzheimer’s using brain -FDG PET (fluorodeoxyglucose positron emission tomography) images, particularly excelling in distinguishing AD from mild cognitive impairment.

CDA-mamba: cross-directional attention mamba for enhanced 3D medical image segmentation

Recent advances in state space models (SSMs) have demonstrated remarkable efficiency in modeling long-range dependencies, yet their application to 3D medical image segmentation remains underexplored. This paper introduces CDA-Mamba (Cross-Directional Attention Mamba), a novel hybrid architecture that combines the efficiency of SSMs with the strengths of convolutional and attention mechanisms to address the unique challenges of 3D medical image segmentation. CDA-Mamba features three key innovations: a Multi-Frequency Gated Convolution (MFGC) module to enhance spatial and frequency-domain feature integration, a Tri-Directional Mamba module to capture volumetric dependencies across orthogonal dimensions, and Selective Self-Attention integration in high-semantic layers to balance computational efficiency with global context modeling. Comprehensive experiments on the BraTS2023 brain tumor segmentation dataset highlight the competitive performance of CDA-Mamba, which achieves an average Dice score of 91.44. Moreover, evaluations on the AIIB2023 airway segmentation dataset further validate its effectiveness, with CDA-Mamba attaining the highest IoU of 88.72 and a DLR of 71.01. These results underscore its ability to balance accuracy and efficiency in 3D medical image segmentation.

Obstructed GNSS Positioning Assisted by Opportunistic LEO Doppler Leveraging Real-Time PPP With State-Space Representation

Obstructed GNSS Positioning Assisted by Opportunistic LEO Doppler Leveraging Real-Time PPP With State-Space Representation

Lightweight real-time speech enhancement: State-space models and multi-spectral scanning techniques

Lightweight real-time speech enhancement: State-space models and multi-spectral scanning techniques

A Lyapunov Function-Based Condition for Nonoscillatory Behaviors in a Class of Memristor Circuits

In recent years, many contributions have been reported in the literature showing that memristor circuits can exhibit very rich oscillatory and chaotic behaviors. In this paper, we look for conditions ensuring that memristor circuits are nonoscillatory, i.e. they do not display oscillations and more complex attractors and enable convergence toward some of the infinite nonisolated equilibrium points. Specifically, we consider the class of memristor circuits composed by the interconnection of a linear time-invariant two-terminal (one port) element and an ideal memristor, which can be either flux-controlled or charge-controlled and whose characteristic satisfies a slope-bounded condition. First, exploiting the well-known fact that any circuit with an ideal memristor admits first integrals, and hence its state space is decomposed in a continuum of invariant manifolds, a state space representation of the circuit dynamics on each invariant manifold is derived in an explicit way. Then, conditions to ensure the absence of oscillatory behaviors on each manifold are obtained by employing the 2-additive compound matrix of the Jacobian of these representations. It is shown that the memristor circuit is nonoscillatory if there exists a common Lyapunov function for two suitable 2-additive compound matrices, which is obtained by solving two linear matrix inequalities. Two memristor circuits are analyzed in some detail to illustrate the application of the results and their degree of conservatism.