arXiv subjects

Xiliang Yang

Publications and source records attributed to Xiliang Yang.

8 recordsLinked to original sources

arXiv2025

Self-Assembled hBN Wrinkles as Planar Optofluidic Channels

Optically accessible, scalable planar nanofluidic channels are attractive for studying transport and localization under confinement. Two dimensional (2D) materials provide large area, atomically flat interfaces for generating such platforms, yet achieving long range one-dimensional (1D) confinement with top-down nanofabrication remains challenging because it requires reproducible nanoscale feature control over extended distances, high yield, and low nonspecific adsorption of analytes under aqueous conditions. Here we demonstrate that thermally induced wrinkling of exfoliated hexagonal boron nitride (hBN) produces self-assembled, liquid-accessible, channel-like networks through a lithography-free process. By varying flake thickness and substrate choice, we quantify statistical trends in wrinkle density and morphology, thereby establishing a practical fabrication design space. Atomic force microscopy and electron microscopy reveal wrinkle-derived geometries with vertical confinement ranging from <2 nm to >100 nm depending on flake thickness and substrate. We further employ time-sequence optical imaging upon droplet-contact, which together with Raman mapping of the water OH-stretch band and capacitance-gradient mapping (dC/dz) by scanning dielectric microscopy (KPFM-based) measurements, demonstrates liquid infiltration and long-term liquid retention within the wrinkle network for more than 10 h. We finally show a proof-of-concept biomolecule confinement application in which we integrate a graphene overlayer as a background suppression interface, enabling wide-field fluorescence localization of ATTO647N labeled DNA along hBN wrinkle-induced nanochannels. Overall, this work establishes self-assembled hBN wrinkles as a scalable, and optically addressable planar nanofluidic platform for confinement of fluids and biomolecules.

physics.app-ph↗

arXiv2025

A Mathematics Framework of Artificial Shifted Population Risk and Its Further Understanding Related to Consistency Regularization

Data augmentation is an important technique in training deep neural networks as it enhances their ability to generalize and remain robust. While data augmentation is commonly used to expand the sample size and act as a consistency regularization term, there is a lack of research on the relationship between them. To address this gap, this paper introduces a more comprehensive mathematical framework for data augmentation. Through this framework, we establish that the expected risk of the shifted population is the sum of the original population risk and a gap term, which can be interpreted as a consistency regularization term. The paper also provides a theoretical understanding of this gap, highlighting its negative effects on the early stages of training. We also propose a method to mitigate these effects. To validate our approach, we conducted experiments using same data augmentation techniques and computing resources under several scenarios, including standard training, out-of-distribution, and imbalanced classification. The results demonstrate that our methods surpass compared methods under all scenarios in terms of generalization ability and convergence stability. We provide our code implementation at the following link: https://github.com/ydlsfhll/ASPR.

cs.LG↗

arXiv2025

DPO-Shift: Shifting the Distribution of Direct Preference Optimization

Direct Preference Optimization (DPO) and its variants have become increasingly popular for aligning language models with human preferences. These methods aim to teach models to better distinguish between chosen (or preferred) and rejected (or dispreferred) responses. However, prior research has identified that the probability of chosen responses often decreases during training, and this phenomenon is known as likelihood displacement. To tackle this challenge, in this work we introduce DPO-Shift to controllably shift the distribution of the chosen probability. Then, we show that DPO-Shift exhibits a fundamental trade-off between improving the chosen probability and sacrificing the reward margin, as supported by both theoretical analysis and experimental validation. Furthermore, we demonstrate the superiority of DPO-Shift over DPO on downstream tasks such as MT-Bench and a designed win rate experiment. We believe this study shows that the likelihood displacement issue of DPO can be effectively mitigated with a simple, theoretically grounded solution. Our code is available at https://github.com/Meaquadddd/DPO-Shift.

cs.CL↗

arXiv2025

An efficient likelihood-free Bayesian inference method based on sequential neural posterior estimation

Sequential neural posterior estimation (SNPE) techniques have been recently proposed for dealing with simulation-based models with intractable likelihoods. Unlike approximate Bayesian computation, SNPE techniques learn the posterior from sequential simulation using neural network-based conditional density estimators by minimizing a specific loss function. The SNPE method proposed by Lueckmann et al. (2017) used a calibration kernel to boost the sample weights around the observed data, resulting in a concentrated loss function. However, the use of calibration kernels may increase the variances of both the empirical loss and its gradient, making the training inefficient. To improve the stability of SNPE, this paper proposes to use an adaptive calibration kernel and several variance reduction techniques. The proposed method greatly speeds up the process of training and provides a better approximation of the posterior than the original SNPE method and some existing competitors as confirmed by numerical experiments. We also managed to demonstrate the superiority of the proposed method for a high-dimensional model with a real-world dataset.

stat.ML↗

arXiv2024

Actuation and mapping of SAW-induced high-frequency wavefields on suspended graphene membranes

High frequency acoustic devices based on two-dimensional (2D) materials are unique platforms to design and manipulate the spatiotemporal response of acoustic waves for next-generation sensing and contactless actuation applications. Conventional methods for actuating suspended membranes, however, cannot be applied to all 2D materials, or are limited in frequency. There is, therefore, a need for a universal high-frequency, on-chip actuation technique that can be applied to all types of membranes. Here, we demonstrate that surface acoustic waves (SAWs) can be used to efficiently actuate suspended 2D materials by exciting suspended graphene membranes with high-frequency (375 MHz) Rayleigh surface waves and mapping the resulting vibration field with atomic force acoustic microscopy (AFAM). Acoustic waves travelling from supported to suspended graphene experience a reduction in acoustic wavelength from 10 μm to ~2 μum due to the decrease in effective bending rigidity, leading to a decrease in wave velocity on suspended graphene. By varying the excitation frequency, we observed a change in phase velocity from ~160 m/s to ~700 m/s. This behavior is consistent with the nonlinear dispersion of acoustic waves, as predicted by plate theory, in suspended graphene membranes. The geometry and bending rigidity of the membrane thus play key roles in modulating the acoustic wave pattern and wavelength. This combined SAW actuation and AFAM visualization scheme can give new insights into the fundamentals of acoustic transport at the nanoscale limit and provides a route towards the manipulation of localized wavefields for on-chip patterning and transport over 2D materials surfaces.

cond-mat.mes-hall↗

arXiv2024

Microsphere-assisted generation of localized optical emitters in 2D hexagonal boron nitride

Crystal defects in hexagonal boron nitride (hBN) are emerging as versatile nanoscale optical probes with a wide application profile, spanning the fields of nanophotonics, biosensing, bioimaging and quantum information processing. However, generating these crystal defects as reliable optical emitters remains challenging due to the need for deterministic defect placement and precise control of the emission area. Here, we demonstrate an approach that integrates microspheres (MS) with hBN optical probes to enhance both defect generation and optical signal readout. This technique harnesses MS to amplify light-matter interactions at the nanoscale through 2 two mechanisms: focused femtosecond (fs) laser irradiation into a photonic nanojet for highly localized defect generation, and enhanced light collection via the whispering gallery mode effect. Our MS-assisted defect generation method reduces the emission area by a factor of 5 and increases the fluorescence collection efficiency by approximately 10 times compared to MS-free samples. These advancements in defect generation precision and signal collection efficiency open new possibilities for optical emitter manipulation in hBN, with potential applications in quantum technologies and nanoscale sensing.

cond-mat.mes-hall↗

arXiv2024

Leveraging Nested MLMC for Sequential Neural Posterior Estimation with Intractable Likelihoods

There is a growing interest in studying sequential neural posterior estimation (SNPE) techniques due to their advantages for simulation-based models with intractable likelihoods. The methods aim to learn the posterior from adaptively proposed simulations using neural network-based conditional density estimators. As an SNPE technique, the automatic posterior transformation (APT) method proposed by Greenberg et al. (2019) performs well and scales to high-dimensional data. However, the APT method requires computing the expectation of the logarithm of an intractable normalizing constant, i.e., a nested expectation. Although atomic proposals were used to render an analytical normalizing constant, it remains challenging to analyze the convergence of learning. In this paper, we reformulate APT as a nested estimation problem. Building on this, we construct several multilevel Monte Carlo (MLMC) estimators for the loss function and its gradients to accommodate different scenarios, including two unbiased estimators, and a biased estimator that trades a small bias for reduced variance and controlled runtime and memory usage. We also provide convergence results of stochastic gradient descent to quantify the interaction of the bias and variance of the gradient estimator. Numerical experiments for approximating complex posteriors with multimodality in moderate dimensions are provided to examine the effectiveness of the proposed methods.

stat.CO↗

arXiv2019

Optimization theory and application of nano-microscopic properties of dielectric microspheres

The dielectric microsphere can be directly embedded into the traditional microscope, which can significantly improve the resolution of the microscope and provide a simple and feasible way to break through the diffraction limit of optical imaging system. However, due to the lack of effective theory and formula system, the resolution limit, magnification and optimal imaging position of the microsphere microsystem cannot be determined, which hinders its application. In this paper, the microscopic theory of dielectric microspheres is studied systematically, and the ray optics is extended to the imaging of dielectric microspheres, so as to establish a formula system containing important optical parameters such as resolution, magnification and imaging position, which provides a solid theoretical basis for further optimization of the nano-microscopic properties of dielectric microspheres. Compared with simulation and experiment, the correctness of the formula system is ensured. The formula shows that the mismatch between refractive index and ambient refractive index limits the resolution of microspheres, and it is an effective method to further improve the resolution to find the optimal refractive index ratio. In addition, the larger refractive index of the medium microsphere and the refractive index of the environment are conducive to the enhancement of the imaging magnification, but the working distance of the microscope objective must be taken into account in the experiment to obtain the ideal imaging position. The theoretical system proposed in this paper effectively explains the optical principle of dielectric microsphere imaging, gives optimized optical parameters, and has been applied in experiments, which is of great value for the practical application of dielectric microsphere nanometer microscopy.

physics.optics↗