Owing to their asymmetric structure, Janus transition metal dichalcogenides (TMDs) provide more flexibility in bandgap tuning than conventional TMDs, with hafnium sulfide selenide (HfSSe) standing out as particularly notable in recent years. This work investigated the material properties of HfSSe and found that it is an indirect bandgap semiconductor material. Subsequently, the liquid-phase-exfoliated HfSSe dispersion was deposited onto a tapered fiber, yielding HfSSe saturable absorber (SA) through evanescent-field enhancement. This SA was integrated into the erbium-doped ring cavity, successfully achieving mode-locking (ML) and Q-switched mode-locking (QSML) pulse output. Under ML operation, the pulse interval and 3 dB bandwidth are 406 ns and 2.43 nm, respectively. The duration of the soliton pulse is 1.49 ps, and the signal-to-noise ratio is approximately 71.3 dB. QSML operation was achieved by adjusting the polarization controller and changing the pump power. In the QSML state, as the pump power was increased from 90 mW to 130 mW, the repetition rate of the pulse envelope rose from 18.04 kHz to 25.76 kHz, while the pulse envelope width decreased from 23.81 µs to 9.32 µs. At pump power of 130 mW, the maximum output power reached 3.21 mW, corresponding to pulse energy of approximately 124.4 nJ. These results clearly demonstrate that HfSSe has significant application potential in the fields of ultrafast photonics, optical communication, and compact pulsed laser sources.
Multimodal pathological diagnosis is a critical paradigm for accurate clinical decision making. Despite substantial progress, existing weakly supervised multimodal pathological diagnostic approaches integrating whole-slide images(WSIs) and structured clinical data typically rely on predefined collaboration strategies that assume uniform modality reliability, failing to account for sample-wise modality-specific noise and quality variations. Under weak supervision, the lack of fine-grained instance-level annotations prevents reliable assessment of sample-wise modality reliability, causing unreliable modalities to be indiscriminately fused and amplified, leading to unstable representation learning and unreliable multimodal decision making. To address this challenge, we propose a confidence-based multimodal diagnostic framework that models sample-wise modality reliability via prediction confidence and incorporates it throughout representation learning and multimodal decision making. Specifically, a dual-path global-local attention mechanism suppresses attention noise and stabilizes confidence estimation in pathological image representations, while large language models(LLMs)-based semantic reasoning improves the quality and consistency of structured clinical data through clinically plausible missing value imputation. At the multimodal decision stage, prediction confidence is used as an indicator of modality reliability to guide adaptive fusion, preventing unreliable modalities from dominating predictions. Evaluated on CAMELYON16, TCGA-BRCA, and TCGA-NSCLC datasets, the proposed model demonstrates robust improvements. On the TCGA-NSCLC dataset, it achieves 93.36% Accuracy and 96.40% AUC, outperforming the best unimodal baseline by 4.86% and 2.11%. Furthermore, it consistently surpasses multimodal approaches with up to 1.42% accuracy improvements, proving its superior robustness under sample-wise modality noise. GitHub repository: https://github.com/KuaLe/Pathologic_Fusion.
In complex and diverse power grid fault scenarios, constructing a systematic fault handling knowledge system is an effective approach to enhancing the intelligence of fault handling. Therefore, we propose a knowledge graph construction method for power grid fault handling based on prompt learning and Low-Rank Adaptation (LoRA) fine-tuning. The proposed method extracts fault handling knowledge from multi-source documents such as fault handling operational regulations and historical cases and represents it in a structured form. The method first segments long documents to satisfy the input length constraint of large language models (LLMs), then designs a stepwise prompting strategy with schema constraints to guide the LoRA-fine-tuned LLM in extracting entities and triples, and finally achieves cross-document knowledge fusion through entity alignment and consistency maintenance to complete knowledge graph construction. Experiments are conducted on four open-source LLMs, namely QwQ-32B, DeepSeek-R1-Distill-Qwen-32B, GLM-4-32B, and gpt-oss-20B, and the results show that QwQ-32B achieves the best performance. After LoRA fine-tuning, the QwQ-32B-tuned model achieves F1 scores of 91.9% and 84.8% for entity and triple extraction, respectively. The constructed knowledge graph can provide decision support for power grid fault handling and serve as a practical reference for LLM-driven knowledge graph construction.
Graph-based electroencephalogram (EEG) emotion recognition methods model inter-channel relationships by constructing connectivity graphs over EEG electrodes. Fixed structural priors may be insufficient to characterize subject- and trial-specific variations, whereas highly adaptive graphs may be sensitive to noise. In addition, temporal and spectral EEG representations are often fused mainly at the feature level, leaving the graph-level relationship between representation domains relatively underexplored. To address these issues, we propose PSAGAN, a Prior-Guided Sample-Adaptive Graph Attention Network with diffusion-based cross-domain graph topology regularization for EEG-based emotion recognition. PSAGAN constructs spatial-spectral and spatial-temporal pathways and introduces a prior-guided sample-adaptive graph attention mechanism within each pathway. This mechanism incorporates distance-based anatomical priors into feature-dependent attention computation, enabling the model to use structural guidance while learning sample-dependent edge weights within anatomically guided neighborhoods. To coordinate the two pathway-specific graphs, PSAGAN introduces a diffusion-based topology regularization strategy that compares multi-scale diffusion signatures and uses category-level contrastive learning to encourage coarse-grained topological coherence while retaining domain-specific learned connectivity patterns. An adaptive integration module then combines the two pathway representations for classification. Experiments on SEED and SEED-IV under both subject-dependent and subject-independent settings, and on DEAP under trial-grouped 10-fold cross-validation, show that PSAGAN achieves competitive performance under the adopted evaluation protocols. On SEED, PSAGAN obtains 98.39% subject-dependent accuracy and 90.18% subject-independent accuracy. On SEED-IV, it achieves 90.57% and 79.36% under the two settings, respectively. On DEAP, it reaches 97.83% and 98.21% for valence and arousal classification, respectively. Ablation studies indicate that prior-guided graph learning, diffusion-based topology regularization, and adaptive integration provide complementary benefits. Additional evaluations on motor imagery, sleep staging, and depression detection suggest potential applicability to broader multi-channel EEG tasks.
Nonlinear optical materials have potential applications in optical signal processing, photovoltaic conversion, and optical information storage. As organic conjugated material, its high second-order nonlinear optical (NLO) properties can be produced due to the introduction of alkali metal species on porphyrin surface. In order to investigate the effects of alkali metal atoms on the second-order NLO properties of porphyrin derivative (PYD) complexes, a series of novel alkali metal porphyrin derivatives (M@PYD and M2F@PYD, M = Li, Na, K) are systematically studied using the density-functional theory calculations. Firstly, the stable structures of all the complexes are confirmed, and the high chemical stability of alkali metal porphyrin complexes is explored by the analysis of binding energy. The nonlinear optical properties of stable structures are analyzed to obtain the dipole moments, polarizabilities and the static first hyperpolarizabilities of the functionalized complexes. The calculated results show that the system has the maximum static first hyperpolarizability (β0 = 2457.573 a.u.) when the Na2F molecule is introduced onto the PYD surface, whereas M@PYD has relatively small static first hyperpolarizability. The present findings will provide the important reference for designing the high-performance porphyrin-based NLO materials.
