arXiv subjects

Zhen Liao

Publications and source records attributed to Zhen Liao.

5 recordsLinked to original sources

arXiv2026

Dual-Strategy-Enhanced ConBiMamba for Neural Speaker Diarization

Conformer and Mamba have achieved strong performance in speech modeling but face limitations in speaker diarization. Mamba is efficient but struggles with local details and nonlinear patterns. Conformer's self-attention incurs high memory overhead for long speech sequences and may cause instability in long-range dependency modeling. These limitations are critical for diarization, which requires both precise modeling of local variations and robust speaker consistency over extended spans. To address these challenges, we first apply ConBiMamba for speaker diarization. We follow the Pyannote pipeline and propose the Dual-Strategy-Enhanced ConBiMamba neural speaker diarization system. ConBiMamba integrates the strengths of Conformer and Mamba, where Conformer's convolutional and feed-forward structures are utilized to improve local feature extraction. By replacing Conformer's self-attention with ExtBiMamba, ConBiMamba efficiently handles long audio sequences while alleviating the high memory cost of self-attention. Furthermore, to address the problem of the higher DER around speaker change points, we introduce the Boundary-Enhanced Transition Loss to enhance the detection of speaker change points. We also propose Layer-wise Feature Aggregation to enhance the utilization of multi-layer representations. The system is evaluated on six diarization datasets and achieves state-of-the-art performance on four of them. The source code of our study is available at https://github.com/lz-hust/DSE-CBM.

cs.SD↗

arXiv2026

Transfer-learned Kolosov-Muskhelishvili Informed Neural Networks for Fracture Mechanics

Physics-informed neural networks have been widely applied to solid mechanics problems. However, balancing the governing partial differential equations and boundary conditions remains challenging, particularly in fracture mechanics, where accurate predictions strongly depend on refined sampling near crack tips. To overcome these limitations, a Kolosov-Muskhelishvili informed neural network with Williams enrichment is developed in this study. Benefiting from the holomorphic representation, the governing equations are satisfied by construction, and only boundary points are required for training. Across a series of benchmark problems, the Kolosov-Muskhelishvili informed neural network shows excellent agreement with analytical and finite element method references, achieving average relative errors below 1\% and $R^2$ above 0.99 for both mode I and mode II loadings. Furthermore, three crack propagation criteria (maximum tangential stress, maximum energy release rate, and principle of local symmetry) are integrated into the framework using a transfer learning strategy to predict crack propagation directions. The predicted paths are nearly identical across all criteria, and the transfer learning strategy reduces the required training time by more than 70\%. Overall, the developed framework provides a unified, mesh-free, and physically consistent approach for accurate and efficient crack propagation analysis.

cs.CE↗

arXiv2022

piRank: A Probabilistic Intent Based Ranking Framework for Facebook Search

While numerous studies have been conducted in the literature exploring different types of machine learning approaches for search ranking, most of them are focused on specific pre-defined problems but only a few of them have studied the ranking framework which can be applied in a commercial search engine in a scalable way. In the meantime, existing ranking models are often optimized for normalized discounted cumulative gains (NDCG) or online click-through rate (CTR), and both types of machine learning models are built based on the assumption that high-quality training data can be easily obtained and well applied to unseen cases. In practice at Facebook search, we observed that our training data for ML models have certain issues. First, tail query intents are hardly covered in our human rating dataset. Second, search click logs are often noisy and hard to clean up due to various reasons. To address the above issues, in this paper, we propose a probabilistic intent based ranking framework (short for piRank), which can: 1) provide a scalable framework to address various ranking issues for different query intents in a divide-and-conquer way; 2) improve system development agility including iteration speed and system debuggability; 3) combine both machine learning and empirical-based algorithmic methods in a systematic way. We conducted extensive experiments and studies on top of Facebook search engine system and validated the effectiveness of this new ranking architecture.

cs.IR↗

arXiv2016

Spoof Plasmon Hybridization

Plasmon hybridization between closely spaced nanoparticles yields new hybrid modes not found in individual constituents, allowing for the engineering of resonance properties and field enhancement capabilities of metallic nanostructure. Experimental verifications of plasmon hybridization have been thus far mostly limited to optical frequencies, as metals cannot support surface plasmons at longer wavelengths. Here, we introduce the concept of 'spoof plasmon hybridization' in highly conductive metal structures and investigate experimentally the interaction of localized surface plasmon resonances (LSPR) in adjacent metal disks corrugated with subwavelength spiral patterns. We show that the hybridization results in the splitting of spoof plasmon modes into bonding and antibonding resonances analogous to molecular orbital rule and plasmonic hybridization in optical spectrum. These hybrid modes can be manipulated to produce enormous field enhancements (larger than 5000) by tuning the separation between disks or alternatively, the disk size, which effectively changes the relative gap size. The impact of the radiation loss is considered to find out the optimum disk size that maximizes field enhancement capabilities. Our investigation not only extends the range of applicability of the hybridization model, but also provides insightful guidance to exporting the exciting applications associated with plasmon hybridization to lower spectral range.

physics.optics↗

arXiv2015

High-order localized spoof surface plasmon resonances and experimental verifications

We theoretically demonstrated and experimentally verified high-order radial spoof localized surface plasmon resonances supported by textured metal particles. Through an effective medium theory and exact numerical simulations, we show the emergence of these geometrically-originated electromagnetic modes at microwave frequencies. The occurrence of high-order radial spoof plasmon resonances is experimentally verified in ultrathin disks. Their spectral and near-field properties are characterized experimentally, showing an excellent agreement with theoretical predictions. Our findings shed light into the nature of spoof localized surface plasmons, and open the way to the design of broadband plasmonic devices able to operate at very different frequency regimes.

physics.plasm-ph↗