arXiv subjects

Yulin Gong

Publications and source records attributed to Yulin Gong.

17 recordsLinked to original sources

arXiv2026

Lightweight ML-Based Automatic Sleep Staging Framework with Constrained CNN and Mamba for Small-Sample EEG Datasets

Automatic sleep staging is a key technology for precise diagnosis and treatment of sleep disorders as well as long-term home sleep monitoring. Portable electroencephalogram (EEG) devices have become the focus of research due to their convenience in data collection. However, current methods still face three major challenges: large parameter sizes that easily lead to overfitting on small datasets, low accuracy in classifying difficult stages such as N1 and REM, unclear optimal training dataset size, and difficulty in deployment. This paper proposes GamSleepNet, a lightweight and low-latency automatic sleep staging framework for single-channel EEG. The framework features the FEB module, which combines improved Gabor kernels with learnable filters for feature extraction, uses the Mamba architecture to build a temporal classification network, introduces a novel contrastive loss and a two-stage training strategy, and experimentally validates the optimal dataset size for single-channel EEG sleep staging models. On the Sleepedf dataset, this model achieves an overall accuracy of 87.86 percent with only 30.86 thousand parameters, with all metrics reaching SOTA levels and significantly improving the identification accuracy of challenging sleep stages.

cs.LG↗

arXiv2026

Full Nonlinear Velocity Reconstruction With Transformer and Ensemble Tree Machine Learning

Accurate reconstruction of peculiar velocities from galaxy positions is important for probing the motion and evolution of large scale structure. They are sensitive to the cosmological effects of gravity and dark sector matter, and complement other velocity inference methods such as the kinematic Sunyaev-Zel'dovich (kSZ) imprinted in the CMB. We show that machine learning methods improve velocity reconstruction by capturing nonlinear contributions. Specifically, we train both a gradient boosting decision tree (GBDT) and a Transformer using multi-scale features to predict the residual between the actual velocity and the estimate from linear theory for both the line-of-sight and transverse components. We evaluate our approach in both periodic box and, more realistic, lightcone settings using mock galaxy catalogs from the \textsc{AbacusSummit} simulations tailored to DESI spectroscopic surveys of luminous red galaxies (LRGs) and emission line galaxies (ELGs). We also assess the impact of redshift uncertainties such as those in Rubin LSST photometry. Both models significantly outperform the linear theory with the Transformer achieving the best performance. They more accurately recover the velocity power spectrum and maintain a higher cross-correlation with the true velocities across a wider range of spatial scales. Finally, we demonstrate two applications relevant to kSZ analyses: estimating cluster pairwise velocity correlations and stacked cluster density profiles. This machine learning framework for nonlinear velocity reconstruction opens up powerful new applications of survey data from DESI, Rubin LSST, Euclid and the Roman Space Telescope.

astro-ph.CO↗

arXiv2026

Observability and Semiclassical Control for Schr\"odinger Equations on Non-compact Hyperbolic Surfaces

We study the observability of the Schr\"odinger equation on $X$, a non-compact covering space of a compact hyperbolic surface $M$. Using a generalized Bloch theory, functions on $X$ are identified as sections of flat Hilbert bundles over $M$. We develop a semiclassical analysis framework for such bundles and generalize the result of semiclassical control estimates in [Dyatlov and Jin, Acta Math., 220 (2018), pp. 297-339] to all flat Hilbert bundles over $M$, with uniform constants with respect to the choice of bundle. Furthermore, when the Riemannian cover $X \to M$ is a normal cover with a virtually Abelian deck transformation group $\Gamma$, we combine the uniform semiclassical control estimates on flat Hilbert bundles with the generalized Bloch theory to derive observability from any $\Gamma$-periodic open subsets of $X$. We also discuss applications of the uniform semiclassical control estimates in spectral geometry.

math.AP↗

arXiv2025

Detection of the Pairwise Kinematic Sunyaev-Zel'dovich Effect and Pairwise Velocity with DESI DR1 Galaxies and ACT DR6 and Planck CMB Data

We present a 9.3-sigma detection of the pairwise kinematic Sunyaev-Zeldovich (kSZ) effect by combining a sample of 913,286 Luminous Red Galaxies (LRGs) from the Dark Energy Spectroscopic Instrument Data Release 1 (DESI DR1) catalog and co-added Atacama Cosmology Telescope (ACT DR6) and Planck cosmic microwave background (CMB) temperature maps. This represents the highest-significance pairwise kSZ measurement to date. The analysis uses three ACT CMB temperature maps: co-added 150 GHz, total frequency maps, and a component-separated Internal Linear Combination (ILC) map, all of which cover 19,000 square degrees of the sky from Advanced ACTPol observations conducted between 2017 and 2022. Comparison of the results of these three maps serves as a consistency check for potential foreground contamination that may depend on the observation frequency. An estimate of the best-fit mass-averaged optical depth is obtained by comparing the pairwise kSZ curve with the linear-theory prediction of the pairwise velocity under the best-fit Planck cosmology, and is compared with predictions from simulations. This estimate serves as a reference point for future comparisons with thermal SZ-derived optical depth measurements for the same DESI cluster samples, which will be presented in a companion paper. Finally, we employ a machine-learning approach trained on simulations to estimate the optical depth for 456,803 DESI LRG-identified clusters within the simulated mass range (greater than about 1e13 solar masses). These are combined with the measured kSZ signal to infer the individual cluster peculiar velocities, providing the opportunity to constrain the behavior of gravity and the dark sector over a range of cosmic scales and epochs.

astro-ph.CO↗

arXiv2025

Iskay2: Signal Extraction of the Kinematic Sunyaev-Zel'dovich Effect Through The Pairwise Estimator. Pipeline and Validation

The peculiar motions of massive halos probe the distribution of matter in the universe, the gravitational potential, and the history of cosmic structure growth. The kinematic Sunyaev-Zeldovich (kSZ) effect offers a robust observational window into these properties. The pairwise kSZ estimator probes the pairwise momentum of groups of galaxies by cross-correlating cosmic microwave background (CMB) maps with spectroscopic galaxy catalogs, using galaxies to trace the positions of dark matter halos. This note introduces iskay2, an efficient pipeline designed to apply the pairwise kSZ estimator to maps of the CMB and large galaxy catalogs. Pairwise kSZ measurements obtained using this pipeline are compared to previously published results and are shown to be consistent within statistical expectations. This pipeline will enable high-precision measurements of the pairwise kSZ utilizing galaxy catalogs like DESI combined with past, current and next-generation high-resolution CMB experiments such as ACT, SPT and the Simons Observatory.

astro-ph.CO↗

arXiv2025

Cluster optical depth and pairwise velocity estimation using machine learning

We apply two machine learning methods, a CNN deep-leaning model and a gradient-boosting decision tree, to estimate individual cluster optical depths from observed properties derived from multiple complementary datasets. The models are trained and tested with simulated N-body derived halo catalogs and synthetic full-sky CMB maps designed to mirror data from the DESI and Simons Observatory experiments. Specifically, the thermal Sunyaev-Zel'dovich (tSZ) and CMB lensing convergence, along with cluster virial mass estimates are used as features to train the machine learning models. The predicted optical depths are combined with kinematic Sunyaev-Zel'dovich (kSZ) measurements to estimate individual cluster radial peculiar velocities. The method is shown to recover an unbiased estimate of the pairwise velocity statistics of the simulated cluster sample. The model's efficacy is demonstrated for halos with mass range $10^{13} M_{\odot} < M_{200} < 10^{15} M_{\odot}$ over a redshift range $0<z<1$, and validated in the presence of primary CMB, instrument noise, lensing convergence noise, and potential uncertainties in halo virial mass estimates. We apply the method to ACT CMB data, using ACT DR4 component-separated maps for tSZ and CMB lensing and ACT DR5 maps for kSZ, in conjunction with galaxy clusters observed in the SDSS DR15 spectroscopic survey. We demonstrate that the machine learning approach is an effective one to analyze data from current and upcoming CMB experiments such as Simons Observatory and CCAT, and galaxy surveys, such as DESI and Roman, for which the pairwise velocity statistics can provide valuable insights into the properties of neutrinos and gravity on cosmic scales.

astro-ph.CO↗

arXiv2025

Sublinear lower bounds of eigenvalues for twisted Laplacian on compact hyperbolic surfaces

We investigate the asymptotic spectral distribution of the twisted Laplacian associated with a real harmonic 1-form on a compact hyperbolic surface. In particular, we establish a sublinear lower bound on the number of eigenvalues in a sufficiently large strip determined by the pressure of the harmonic 1-form. Furthermore, following an observation by Anantharaman \cite{nalinideviation}, we show that quantum unique ergodicity fails to hold for certain twisted Laplacians.

math.SP↗

arXiv2025

The Atacama Cosmology Telescope: DR6 Maps

We present Atacama Cosmology Telescope (ACT) Data Release 6 (DR6) maps of the Cosmic Microwave Background temperature and polarization anisotropy at arcminute resolution over three frequency bands centered on 98, 150 and 220 GHz. The maps are based on data collected with the AdvancedACT camera over the period 2017--2022 and cover 19,000 square degrees with a median combined depth of 10 uK arcmin. We describe the instrument, mapmaking and map properties and illustrate them with a number of figures and tables. The ACT DR6 maps and derived products are available on LAMBDA at https://lambda.gsfc.nasa.gov/product/act/actadv_prod_table.html. We also provide an interactive web atlas at https://phy-act1.princeton.edu/public/snaess/actpol/dr6/atlas and HiPS data sets in Aladin (e.g. https://alasky.cds.unistra.fr/ACT/DR4DR6/color_CMB).

astro-ph.CO↗

arXiv2025

The Atacama Cosmology Telescope: DR6 Power Spectra, Likelihoods and $\Lambda$CDM Parameters

We present power spectra of the cosmic microwave background (CMB) anisotropy in temperature and polarization, measured from the Data Release 6 maps made from Atacama Cosmology Telescope (ACT) data. These cover 19,000 deg$^2$ of sky in bands centered at 98, 150 and 220 GHz, with white noise levels three times lower than Planck in polarization. We find that the ACT angular power spectra estimated over 10,000 deg$^2$, and measured to arcminute scales in TT, TE and EE, are well fit by the sum of CMB and foregrounds, where the CMB spectra are described by the $\Lambda$CDM model. Combining ACT with larger-scale Planck data, the joint P-ACT dataset provides tight limits on the ingredients, expansion rate, and initial conditions of the universe. We find similar constraining power, and consistent results, from either the Planck power spectra or from ACT combined with WMAP data, as well as from either temperature or polarization in the joint P-ACT dataset. When combined with CMB lensing from ACT and Planck, and baryon acoustic oscillation data from DESI DR1, we measure a baryon density of $\Omega_b h^2=0.0226\pm0.0001$, a cold dark matter density of $\Omega_c h^2=0.118\pm0.001$, a Hubble constant of $H_0=68.22\pm0.36$ km/s/Mpc, a spectral index of $n_s=0.974\pm0.003$, and an amplitude of density fluctuations of $\sigma_8=0.813\pm0.005$. Including the DESI DR2 data tightens the Hubble constant to $H_0=68.43\pm0.27$ km/s/Mpc; $\Lambda$CDM parameters agree between the P-ACT and DESI DR2 data at the $1.6\sigma$ level. We find no evidence for excess lensing in the power spectrum, and no departure from spatial flatness. The contribution from Sunyaev-Zel'dovich (SZ) anisotropy is detected at high significance; we find evidence for a tilt with suppressed small-scale power compared to our baseline SZ template spectrum, consistent with hydrodynamical simulations with feedback.

astro-ph.CO↗

arXiv2025

The Atacama Cosmology Telescope: DR6 Constraints on Extended Cosmological Models

We use new cosmic microwave background (CMB) primary temperature and polarization anisotropy measurements from the Atacama Cosmology Telescope (ACT) Data Release 6 (DR6) to test foundational assumptions of the standard cosmological model and set constraints on extensions to it. We derive constraints from the ACT DR6 power spectra alone, as well as in combination with legacy data from Planck. To break geometric degeneracies, we include ACT and Planck CMB lensing data and baryon acoustic oscillation data from DESI Year-1, and further add supernovae measurements from Pantheon+ for models that affect the late-time expansion history. We verify the near-scale-invariance (running of the spectral index $d n_s/d\ln k = 0.0062 \pm 0.0052$) and adiabaticity of the primordial perturbations. Neutrino properties are consistent with Standard Model predictions: we find no evidence for new light, relativistic species that are free-streaming ($N_{\rm eff} = 2.86 \pm 0.13$, which combined with external BBN data becomes $N_{\rm eff} = 2.89 \pm 0.11$), for non-zero neutrino masses ($\sum m_\nu < 0.082$ eV at 95% CL), or for neutrino self-interactions. We also find no evidence for self-interacting dark radiation ($N_{\rm idr} < 0.134$), early-universe variation of fundamental constants, early dark energy, primordial magnetic fields, or modified recombination. Our data are consistent with standard BBN, the FIRAS-inferred CMB temperature, a dark matter component that is collisionless and with only a small fraction allowed as axion-like particles, a cosmological constant, and the late-time growth rate predicted by general relativity. We find no statistically significant preference for a departure from the baseline $\Lambda$CDM model. In general, models introduced to increase the Hubble constant or to decrease the amplitude of density fluctuations inferred from the primary CMB are not favored by our data.

astro-ph.CO↗

arXiv2025

Measurements of the Thermal Sunyaev-Zel'dovich Effect with ACT and DESI Luminous Red Galaxies

Cosmic Microwave Background (CMB) photons scatter off the free-electron gas in galaxies and clusters, allowing us to use the CMB as a backlight to probe the gas in and around low-redshift galaxies. The thermal Sunyaev-Zel'dovich effect, sourced by hot electrons in high-density environments, measures the thermal pressure of the target objects, shedding light on halo thermodynamics and galaxy formation and providing a path toward understanding the baryon distribution around cosmic structures. We use a combination of high-resolution CMB maps from the Atacama Cosmology Telescope (ACT) and photometric luminous red galaxy (LRG) catalogues from the Dark Energy Spectroscopic Instrument (DESI) to measure the thermal Sunyaev-Zel'dovich signal in four redshift bins from $z=0.4$ to $z=1.2$, with a combined detection significance of 19$\sigma$ when stacking on the fiducial CMB Compton-$y$ map. We discuss possible sources of contamination, finding that residual dust emission associated with the target galaxies is important and limits current analyses. We discuss several mitigation strategies and quantify the residual modelling uncertainty. This work complements closely related measurements of the kinematic Sunyaev-Zel'dovich and weak lensing of the same galaxies.

astro-ph.CO↗

arXiv2024

The spectral concentration for damped waves on compact Anosov manifolds

We study the spectral distribution of damped waves on compact Anosov manifolds. Sj\"ostrand \cite{SJ1} proved that the imaginary parts of the majority of the eigenvalues concentrate near the average of the damping function, see also Anantharaman \cite{AN2}. In this paper, we prove that the most of eigenvalues actually lie in certain regions with imaginary parts that approach the average logarithmically as the real parts tend to infinity. The proof relies on the moderate deviation principles for Anosov geodesic flows. As an application, we show the concentration of non-trivial zeros of twisted Selberg zeta functions in a logarithmic region asymptotically close to $\Re s=\frac{1}{2}$.

math.SP↗

arXiv2024

Discrete trace formulas and holomorphic functional calculus for the adjacency matrix of regular graphs

We provide a unified method to study the adjacency matrices of regular graphs (including infinite ones) using holomorphic functional calculus. By applying this calculus on a specific ellipse that contains the spectrum, we derive an expansion of $h(A)$ using non-backtracking matrices. This framework allows us to systematically obtain discrete trace formulas that link spectral theory with graph combinatorics. To show how this method works, we give new proofs for several well-known problems, such as walk counting, the Ihara-Bass formula, and solutions to the heat and Schr\"odinger equations on graphs.

math.CO↗

arXiv2024

Spectral convergence of random regular graphs: Chebyshev polynomials, non-backtracking walks, and unitary-color extensions

In this paper, we give a short proof of the weak convergence to the Kesten-McKay distribution for the normalized spectral measures of random $N$-lifts. This result is derived by generalizing a formula of Friedman involving Chebyshev polynomials and non-backtracking walks. We also extend a criterion of Sodin on the convergence of graph spectral measures to regular graphs of growing degree. As a result, we show that for a sequence of random $(q_n+1)$-regular graphs $G_n$ with $n$ vertices, if $q_n = n^{o(1)}$ and $q_n$ tends to infinity, the normalized spectral measure converges almost surely in $p$-Wasserstein distance to the semicircle distribution for any $p \in [1, \infty)$. This strengthens a result of Dumitriu and Pal. Many of the results are extended to unitary-colored regular graphs.

math.CO↗

arXiv2024

kSZ Pairwise Velocity Reconstruction with Machine Learning

We demonstrate that pairwise peculiar velocity correlations for galaxy clusters can be directly reconstructed from the kinematic Sunyaev-Zel'dovich (kSZ) signature imprinted in the CMB using a machine learning model with a gradient boosting algorithm trained on high-fidelity kSZ simulations. The machine learning model is trained using six to eight cluster features that are directly related to observables from CMB and large-scale structure surveys. We validate the capabilities of the approach in light of the presence of primary CMB, detector noise, and potential uncertainties in the cluster mass estimate and cluster center location. The pairwise velocity statistics extracted using the techniques developed here have the potential to elicit valuable cosmological constraints on dark energy, modified gravity models, and massive neutrinos with kSZ measurements from upcoming CMB surveys, including the Simons Observatory, CMB-S4 and CCAT, and the DESI and SDSS galaxy surveys.

astro-ph.CO↗

arXiv2024

Pairwise kSZ signal extraction efficacy and optical depth estimation

We determine the efficacy of the kinematic Sunyaev-Zel'dovich signal extraction pipeline, using pairwise kSZ measurements, in recovering unbiased estimates of the signal and inference of the associated optical depth. We consider the impact of cluster co-alignments along the line of sight, the modeling of baryonic clustering, and the presence of diffuse gas, as well as instrument beam convolution and noise. We demonstrate that two complementary approaches, aperture photometry, and a matched filter, can be used to recover an unbiased estimate of the cluster kSZ signal and the associated optical depth. Aperture photometry requires a correction factor accounting for the subtraction of signal in the annulus while the matched filter requires a tuning of the signal template profile. We show that both of these can be calibrated from simulated survey data. The optical depth estimates are also consistent with those inferred from stacked thermal SZ measurements. We apply the approaches to the publicly available Atacama Cosmology Telescope (ACT) data. The techniques developed here provide a promising method to leverage upcoming kSZ measurements, from ACT, Simons Observatory, CCAT, and CMB-S4 with spectroscopic galaxy surveys from DESI, Euclid, and Roman, to constrain cosmological properties of the dark energy, gravity, and neutrino masses.

astro-ph.CO↗

arXiv2023

Spectral Distribution of Twisted Laplacian on Typical Hyperbolic Surfaces of High Genus

We investigate the spectral distribution of the twisted Laplacian associated with uniform square-integrable bounded harmonic 1-form on typical hyperbolic surfaces of high genus. First, we estimate the spectral distribution by the supremum norm of the corresponding harmonic form. Subsequently, we show that the square-integrable bounded harmonic form exhibits a small supremum norm for typical hyperbolic surfaces of high genus. Based on these findings, we prove a uniform Weyl law for the distribution of real parts of the spectrum on typical hyperbolic surfaces.

math.SP↗