Hyperspectral target detection (HTD) identifies objects of interest from complex backgrounds at the pixel level, playing a vital role in Earth observation. However, the limited target priors constrain the ability to obtain sufficient features or patterns for background-target discrimination, and spectral variation further exacerbates the difficulty of achieving reliable and robust performance. To address these challenges, this paper proposes an efficient self-supervised HTD method with a pyramid state space model (SSM), named HTD-Mamba, which employs spectrally contrastive learning to distinguish between target and background based on the similarity measurement of intrinsic features. Specifically, to obtain sufficient training samples and leverage spatial contextual information, we propose a spatial-encoded spectral augmentation technique that encodes all surrounding pixels within a patch into a transformed view of the center pixel. Additionally, to explore global band correlations, we divide pixels into continuous group-wise spectral embeddings and introduce Mamba to HTD for the first time to model long-range dependencies of the spectral sequence with linear complexity. Furthermore, to alleviate spectral variation and enhance robust representation, we propose a pyramid SSM as a backbone to capture and fuse multiresolution spectral-wise intrinsic features. Extensive experiments conducted on four public datasets demonstrate that the proposed method outperforms state-of-the-art methods in both quantitative and qualitative evaluations. Code is available at https://github.com/shendb2022/HTD-Mamba.

Large language models (LLMs) have shown impressive general-purpose language capabilities, but their application in specialized domains such as healthcare and law remains limited due to two major challenges, namely, a lack of deep domain-specific knowledge and the inability to incorporate real-time information updates. This paper focuses on addressing these challenges by introducing parameter-sensitive low-rank adaptation (LoRA) and retrieval-augmented generation (RAG), named SensiLoRA-RAG, a two-stage framework designed to enhance LLM performance in domain-specific question-answering tasks. In the first stage, we propose a parameter-sensitive LoRA fine-tuning method that efficiently adapts LLMs to specialized domains using limited high-quality professional data, enabling rapid and resource-efficient specialization. In the second stage, we develop a chain-of-thought RAG mechanism that dynamically retrieves and integrates up-to-date external knowledge, improving the model’s ability to reason with current information and complex domain context. We evaluate our framework on tasks in the medical and legal fields, demonstrating that SensiLoRA-RAG significantly improves answer accuracy, domain relevance, and adaptability compared to baseline methods.

Time series forecasting (TSF) has extensive applications in domains, such as energy, traffic, and weather prediction. Currently, existing literature has designed many architectures that combine deep learning models in the frequency domain, and effective results have been achieved. However, handling complex-valued arithmetic poses a challenge for most frequency domain-based models. Additionally, features extracted solely in either the time or frequency domain are not comprehensive enough. To solve these problems, we propose a fully complex network (FCNet) in this work, where all network layers are adapted to handle complex-valued computations to simultaneously learn the information in the real and imaginary parts. First, we utilize time-frequency conversion to obtain time-frequency domain signals. And then we design the time-frequency filter-enhanced block to effectively capture global features from time-frequency signals. Finally, we design the complex-valued time-frequency Transformers Block, which separately extracts information from the time and frequency domains. Experimental evaluations on eight data sets from five benchmark domains demonstrate that our model significantly outperforms state-of-the-art methods in TSF. Code is available at https://github.com/ZHU-0108/FCNet-main .

Multivariate time series data provide a robust framework for future predictions by leveraging information across multiple dimensions, ensuring broad applicability in practical scenarios. However, their high dimensionality and mixing patterns pose significant challenges in establishing an interpretable and explicit mapping between historical and future series, as well as extracting long-range feature dependencies. To address these challenges, we propose a channel-time dual unmixing network for multivariate time series forecasting (named MTS-UNMixer), which decomposes the entire series into critical bases and coefficients across both the time and channel dimensions. This approach establishes a robust sharing mechanism between historical and future series, enabling accurate representation and enhancing physical interpretability. Specifically, MTS-UNMixers represent sequences over time as a mixture of multiple trends and cycles, with the time-correlated representation coefficients shared across both historical and future time periods. In contrast, sequence over channels can be decomposed into multiple tick-wise bases, which characterize the channel correlations and are shared across the whole series. To estimate the shared time-dependent coefficients, a vanilla Mamba network is employed, leveraging its alignment with directional causality. Conversely, a bidirectional Mamba network is utilized to model the shared channel-correlated bases, accommodating noncausal relationships. Experimental results show that MTS-UNMixers significantly outperform existing methods on multiple benchmark datasets. The code is available at https://github.com/ZHU-0108/MTS-UNMixers.