arXiv subjects

Yu-Hsiang Cheng

Publications and source records attributed to Yu-Hsiang Cheng.

9 recordsLinked to original sources

arXiv2026

Testing for cross-quantilogram change

For two time series $\{ (Y_t, Z_t^Y) \}_{t}$ and $\{(X_t, Z_t^X)\}_{t}$, the directional dependence of $\{ X_t \}_{t}$ on $\{ Y_t \}_{t}$ while removing the impact of $Z_t^X$ on $X_t$ and the impact of $Z_t^Y$ on $ Y_t$ can be measured by cross-quantilograms. When the two time series are obeserved over two periods of time, it can be of interest to learn whether the cross-quantilograms remain the same for the two periods of time. We propose a test for this purpose, and the cross-quantilograms are estimated using the estimators proposed by Han (2016). The $p$-value of the proposed test is obtained based on a bootstrap approach.

stat.ME↗

arXiv2023

A mixture logistic model for panel data with a Markov structure

In this study, we propose a mixture logistic regression model with a Markov structure, and consider the estimation of model parameters using maximum likelihood estimation. We also provide a forward type variable selection algorithm to choose the important explanatory variables to reduce the number of parameters in the proposed model.

stat.ME↗

arXiv2022

LG-LSQ: Learned Gradient Linear Symmetric Quantization

Deep neural networks with lower precision weights and operations at inference time have advantages in terms of the cost of memory space and accelerator power. The main challenge associated with the quantization algorithm is maintaining accuracy at low bit-widths. We propose learned gradient linear symmetric quantization (LG-LSQ) as a method for quantizing weights and activation functions to low bit-widths with high accuracy in integer neural network processors. First, we introduce the scaling simulated gradient (SSG) method for determining the appropriate gradient for the scaling factor of the linear quantizer during the training process. Second, we introduce the arctangent soft round (ASR) method, which differs from the straight-through estimator (STE) method in its ability to prevent the gradient from becoming zero, thereby solving the discrete problem caused by the rounding process. Finally, to bridge the gap between full-precision and low-bit quantization networks, we propose the minimize discretization error (MDE) method to determine an accurate gradient in backpropagation. The ASR+MDE method is a simple alternative to the STE method and is practical for use in different uniform quantization methods. In our evaluation, the proposed quantizer achieved full-precision baseline accuracy in various 3-bit networks, including ResNet18, ResNet34, and ResNet50, and an accuracy drop of less than 1% in the quantization of 4-bit weights and 4-bit activations in lightweight models such as MobileNetV2 and ShuffleNetV2.

cs.CV↗

arXiv2022

Snapshots of a light-induced metastable hidden phase driven by the collapse of charge order

Nonequilibrium hidden states, both transient and long-lived, provide a unique window into thermally inaccessible regimes of strong coupling between microscopic degrees of freedom in quantum materials. Understanding the physical origin of these states is of both fundamental and practical significance, allowing the exploration of far-from-equilibrium thermodynamics and the development of optoelectronic devices with on-demand photoresponses. However, mapping the ultrafast formation of a long-lived hidden phase remains a long-standing challenge in physics since the initial state of the system is not recovered rapidly and conventional pump-probe methods are thus not applicable. Here, using a suite of state-of-the-art single-shot spectroscopy techniques, we present a direct ultrafast visualization of the photoinduced phase transition to both transient and long-lived hidden states in an electronic crystal, 1T-TaS2. Capturing the dynamics of this complex phase transformation in a single-shot fashion demonstrates a commonality in microscopic pathways, driven by the collapse of charge order, that the system undergoes to enter the hidden state and provides unambiguous spectral fingerprints that distinguish such state from thermally accessible phases. We present a theory of fluctuation-dominated process that explains both the dynamics and the nature of the metastable state. Our results settle the debate around the origin of this elusive metastable state and pave the way for the discovery of new quantum phases of matter.

cond-mat.str-el↗

arXiv2019

Dynamics of a Persistent Insulator-to-Metal Transition in Strained Manganite Films

Transition metal oxides possess complex free energy surfaces with competing degrees of freedom. Photoexcitation allows shaping of such rich energy landscapes. In epitaxially strained $\mathrm{La_{0.67}Ca_{0.33}MnO_3}$, optical excitation with a sub-100 fs pulse above $2\ \mathrm{mJ/cm^2}$ leads to a persistent metallic phase below 100 K. Using single-shot optical and terahertz spectroscopy, we show that this phase transition is a multi-step process. We conclude that the phase transition is driven by partial charge order melting, followed by growth of the persistent metallic phase on longer timescales. A time-dependent Ginzburg-Landau model can describe the fast dynamics of the reflectivity, followed by longer timescale in-growth of the metallic phase.

cond-mat.str-el↗

arXiv2019

Random Partitioning and Distribution-based Thresholding for Iterative Variable Screening in High Dimensions

In big data analysis, a simple task such as linear regression can become very challenging as the variable dimension $p$ grows. As a result, variable screening is inevitable in many scientific studies. In recent years, randomized algorithms have become a new trend and are playing an increasingly important role for large scale data analysis. In this article, we combine the ideas of variable screening and random partitioning to propose a new iterative variable screening method. For moderate sized $p$ of order $O(n^{2-δ})$, we propose a basic algorithm that adopts a distribution-based thresholding rule. For very large $p$, we further propose a two-stage procedure. This two-stage procedure first performs a random partitioning to divide predictors into subsets of manageable size of order $O(n^{2-δ})$ for variable screening, where $δ>0$ can be an arbitrarily small positive number. Random partitioning is repeated a few times. Next, the final estimate of variable subset is obtained by integrating results obtained from multiple random partitions. Simulation studies show that our method works well and outperforms some renowned competitors. Real data applications are presented. Our algorithms are able to handle predictors in the size of millions.

stat.ME↗

arXiv2018

A clustering method for misaligned curves

We consider the problem of clustering misaligned curves. According to our similarity measure, two curves are considered similar if they have the same shape after being aligned, and the warping function does not differ from the identity function very much. A clustering method is proposed, which updates curves so that similar curves become more similar, and then combines curves that are similar enough to form clusters. The proposed method needs to be used together with a clustering index and a set of combination thresholds. Simulation results are presented to demonstrate the performance of this approach under different parameter settings and clustering indexes. Two real data applications are included.

stat.ME↗

arXiv2017

Real-time observation of a coherent lattice transformation into a high-symmetry phase

Excursions far from their equilibrium structures can bring crystalline solids through collective transformations including transitions into new phases that may be transient or long-lived. Direct spectroscopic observation of far-from-equilibrium rearrangements provides fundamental mechanistic insight into chemical and structural transformations, and a potential route to practical applications, including ultrafast optical control over material structure and properties. However, in many cases photoinduced transitions are irreversible or only slowly reversible, or the light fluence required exceeds material damage thresholds. This precludes conventional ultrafast spectroscopy in which optical excitation and probe pulses irradiate the sample many times, each measurement providing information about the sample response at just one probe delay time following excitation, with each measurement at a high repetition rate and with the sample fully recovering its initial state in between measurements. Using a single-shot, real-time measurement method, we were able to observe the photoinduced phase transition from the semimetallic, low-symmetry phase of crystalline bismuth into a high-symmetry phase whose existence at high electronic excitation densities was predicted based on earlier measurements at moderate excitation densities below the damage threshold. Our observations indicate that coherent lattice vibrational motion launched upon photoexcitation with an incident fluence above 10 mJ/cm2 in bulk bismuth brings the lattice structure directly into the high-symmetry configuration for tens of picoseconds, after which carrier relaxation and diffusion restore the equilibrium lattice configuration.

cond-mat.mtrl-sci↗

arXiv2016

A nonparametric copula density estimator incorporating information on bivariate marginals

We propose a copula density estimator that can include information on bivariate marginals when the information is available. We use B-splines for copula density approximation and include information on bivariate marginals via a penalty term. Our estimator satisfies the constraints for a copula density. Under mild conditions, the proposed estimator is consistent.

stat.ME↗