arXiv subjects

Qing Shen

Publications and source records attributed to Qing Shen.

10 recordsLinked to original sources

arXiv2026

RooflineBench: A Benchmarking Framework for On-Device LLMs via Roofline Analysis

The transition toward localized intelligence through Small Language Models (SLMs) has intensified the need for rigorous performance characterization on resource-constrained edge hardware. However, objectively measuring the theoretical performance ceilings of diverse architectures across heterogeneous platforms remains a formidable challenge. In this work, we propose a systematic framework based on the Roofline model that unifies architectural primitives and hardware constraints through the lens of operational intensity (OI). By defining an inference-potential region, we introduce the Relative Inference Potential as a novel metric to compare efficiency differences between Large Language Models (LLMs) on the same hardware substrate. Extensive empirical analysis across diverse compute tiers reveals that variations in performance and OI are significantly influenced by sequence length. We further identify a critical regression in OI as model depth increases. Additionally, our findings highlight an efficiency trap induced by hardware heterogeneity and demonstrate how structural refinements, such as Multi-head Latent Attention (MLA), can effectively unlock latent inference potential across various hardware substrates. These insights provide actionable directions for hardware-software co-design to align neural structures with physical constraints in on-device intelligence. The released code is available in the Appendix C.

cs.LG↗

arXiv2025

Logical Structure as Knowledge: Enhancing LLM Reasoning via Structured Logical Knowledge Density Estimation

The reasoning capabilities of Large Language Models (LLMs) are increasingly attributed to training data quality rather than mere parameter scaling. However, existing data-centric paradigms often equate quality with factuality or diversity and ignore the internal logical complexity of training samples. In this work, we propose that natural language harbors Structured Logical Knowledge manifested through entailment relationships and logical topologies. To quantify this, we introduce Structured Logical Knowledge Density (SLKD), a novel metric that measures logical information content by decomposing natural language into executable predicates and logical primitives. Our analysis reveals a significant logical disparity in current datasets where sparse logical signals predominate. Consequently, we propose a density aware re-cognizing optimization strategy that prioritizes high-density logical samples to enhance with the LLM's reasoning ability. Extensive experiments demonstrate that our approach enhances reasoning performance and generalization without increasing total data volume. These results, further validated within a reinforcement learning framework, suggest that elevating logical density is more critical than expanding data scale for realizing the full cognitive potential of LLMs. The released code is available in the Appendix C.

cs.AI↗

arXiv2024

Physics-Informed AI Inverter

This letter devises an AI-Inverter that pilots the use of a physics-informed neural network (PINN) to enable AI-based electromagnetic transient simulations (EMT) of grid-forming inverters. The contributions are threefold: (1) A PINN-enabled AI-Inverter is formulated; (2) An enhanced learning strategy, balanced-adaptive PINN, is devised; (3) extensive validations and comparative analysis of the accuracy and efficiency of AI-Inverter are made to show its superiority over the classical electromagnetic transient programs (EMTP).

eess.SY↗

arXiv2023

Physics-Aware Neural Dynamic Equivalence of Power Systems

This letter devises Neural Dynamic Equivalence (NeuDyE), which explores physics-aware machine learning and neural-ordinary-differential-equations (ODE-Net) to discover a dynamic equivalence of external power grids while preserving its dynamic behaviors after disturbances. The contributions are threefold: (1) an ODE-Net-enabled NeuDyE formulation to enable a continuous-time, data-driven dynamic equivalence of power systems; (2) a physics-informed NeuDyE learning method (PI-NeuDyE) to actively control the closed-loop accuracy of NeuDyE without an additional verification module; (3) a physics-guided NeuDyE (PG-NeuDyE) to enhance the method's applicability even in the absence of analytical physics models. Extensive case studies in the NPCC system validate the efficacy of NeuDyE, and, in particular, its capability under various contingencies.

eess.SY↗

arXiv2023

Physics-Informed Induction Machine Modelling

This rapid communication devises a Neural Induction Machine (NeuIM) model, which pilots the use of physics-informed machine learning to enable AI-based electromagnetic transient simulations. The contributions are threefold: (1) a formation of NeuIM to represent the induction machine in phase domain; (2) a physics-informed neural network capable of capturing fast and slow IM dynamics even in the absence of data; and (3) a data-physics-integrated hybrid NeuIM approach which is adaptive to various levels of data availability. Extensive case studies validate the efficacy of NeuIM and in particular, its advantage over purely data-driven approaches.

cs.LG↗

arXiv2023

Scalable Neural Dynamic Equivalence for Power Systems

Traditional grid analytics are model-based, relying strongly on accurate models of power systems, especially the dynamic models of generators, controllers, loads and other dynamic components. However, acquiring thorough power system models can be impractical in real operation due to inaccessible system parameters and privacy of consumers, which necessitate data-driven dynamic equivalencing of unknown subsystems. Learning reliable dynamic equivalent models for the external systems from SCADA and PMU data, however, is a long-standing intractable problem in power system analysis due to complicated nonlinearity and unforeseeable dynamic modes of power systems. This paper advances a practical application of neural dynamic equivalence (NeuDyE) called Driving Port NeuDyE (DP-NeuDyE), which exploits physics-informed machine learning and neural-ordinary-differential-equations (ODE-NET) to discover a dynamic equivalence of external power grids while preserving its dynamic behaviors after disturbances. The new contributions are threefold: A NeuDyE formulation to enable a continuous-time, data-driven dynamic equivalence of power systems, saving the effort and expense of acquiring inaccessible system; An introduction of a Physics-Informed NeuDyE learning (PI-NeuDyE) to actively control the closed-loop accuracy of NeuDyE; and A DP-NeuDyE to reduce the number of inputs required for the training. We conduct extensive case studies on the NPCC system to validate the generalizability and accuracy of both PI-NeuDyE and DP-NeuDyE, which span a multitude of scenarios, differing in the time required for fault clearance, the specific fault locations, and the limitations of data. Test results have demonstrated the scalability and practicality of NeuDyE, showing its potential to be used in ISO and utility control centers for online transient stability analysis and for planning purposes.

eess.SY↗

arXiv2022

Efficient Extraction of Hot Carriers in Perovskite Quantum Dot through Building State Coupled Complex

Utilizing hot carriers is the crucial approach for solar cell to exceed the thermodynamic detailed balance limit, yet effective extraction of hot carriers in absorber materials via most commonly used semiconductor acceptors has been a challenge in both materials and photophysics research for many years. Herein, we build series of CsPbI3 quantum dot and fullerene derivative systems to explore the decisive factors of this process and have for the first time realized efficient hot carrier extraction in these systems (maximum extraction efficiency ~ 84%). We find building the systems as state-coupled complexes creates new carrier transport channels at about 0.22 eV above CsPbI3 quantum dot bandgap, which facilitates highly efficient HC extraction. Our research directly visualizes the inner connection of molecule interaction and ultrafast hot carrier extraction. The knowledge and strategy gained here are of universal meaning, taking an important step forward true hot carrier photovoltaics.

physics.app-ph↗

arXiv2019

Data augmentation in microscopic images for material data mining

Recent progress in material data mining has been driven by high-capacity models trained on large datasets. However, collecting experimental data (real data) has been extremely costly since the amount of human effort and expertise required. Here, we develop a novel transfer learning strategy to address small or insufficient data problem. This strategy realizes the fusion of real and simulated data, and the augmentation of training data in data mining procedure. For a specific task of image segmentation, this strategy can generate synthetic images by fusing physical mechanism of simulated images and "image style" of real images. The result shows that the model trained with the acquired synthetic images and 35% of the real images outperforms the model trained on all real images. As the time required to generate synthetic data is almost negligible, this strategy is able to reduce the time cost of real data preparation by roughly 65%.

cond-mat.mtrl-sci↗

arXiv2017

Thinned Coprime Arrays for DOA Estimation

Sparse arrays can generate a larger aperture than traditional uniform linear arrays (ULA) and offer enhanced degrees-of-freedom (DOFs) which can be exploited in both beamforming and direction-of-arrival (DOA) estimation. One class of sparse arrays is the coprime array, composed of two uniform linear subarrays which yield an effective difference co-array with higher number of DOFs. In this work, we present a new coprime array structure termed thinned coprime array (TCA), which exploits the redundancy in the structure of the existing coprime array and achieves the same virtual aperture and DOFs as the conventional coprime array with much fewer number of sensors. An analysis of the DOFs provided by the new structure in comparison with other sparse arrays is provided and simulation results for DOA estimation using the compressive sensing based method are provided.

cs.IT↗

arXiv2016

End-to-End Architecture Modularisation and Slicing for Next Generation Networks

The journey towards the deployment of next generation networks has recently accelerated, driven by the joint effort of research and standards organisations. Despite this fact, the overall picture is still unclear as prioritization and understanding on several key concepts are not yet agreed by major vendors and network providers. Network Slicing is one of the central topics of the debate, and it is expected to become the key feature of next generation networks, providing the flexibility required to support the variety of 5G use cases and business. Network slices are seen as network operator business, offering the possibility to provide flexible services and even infrastructures to vertical industries and classical Telco customers alike. Another key ingredient is the Architecture Modularisation concept, discussed in this paper and regarded by the authors as the essential design principle to build a flexible network architecture natively supporting Network Slicing. According to this concept, conventional monolithic network functions, often corresponding to physical network elements in the existing systems, are to split into basic building blocks defined with the proper granularity, allowing the definition of different logical architectures (i.e. different Network Slices). In this paper, we further discuss a modularisation methodology as a criteria to define the right set of basic building blocks. Defined through this proposed methodology, the set of basic building blocks and the relating interfacing model are discussed. The paper concludes by proposing a modular 5G network architecture as candidate for next generation network standards.

cs.NI↗