arXiv subjects

Jiashu Han

Publications and source records attributed to Jiashu Han.

9 recordsLinked to original sources

arXiv2026

A Marine Debris Detection Framework for Ocean Robots via Self-Attention Enhancement and Feature Interaction Optimization

Marine debris detection for ocean robot is crucial for ecological protection, yet performance is often degraded by low-quality images with blur, complex backgrounds, and small targets. To address these challenges, we propose YOLO-MD, an enhanced YOLO-based detection framework. A Dual-Branch Convolutional Enhanced Self-Attention (DB-CASA) module is designed to strengthen spatial-channel interactions, improving feature representation in degraded images. Additionally, a lightweight shift-based operation is introduced to enhance fine-grained feature extraction for objects of varying scales while maintaining parameter efficiency. We further propose SFG-Loss to mitigate class imbalance and optimization instability via dynamic sample reweighting. Experiments on the UODM dataset demonstrate that YOLO-MD achieves 0.875 precision, 0.822 F1-score, and 0.849 mAP50, outperforming the latest state-of-the-art methods. The effectiveness of this method has also been verified through real-world robotic edge deployment experiments.

cs.CV↗

arXiv2026

MAD: Microenvironment-Aware Distillation -- A Pretraining Strategy for Virtual Spatial Omics from Microscopy

Bridging microscopy and omics would allow us to read molecular states from images-at single-cell resolution and tissue scale-without the cost and throughput limits of omics technologies. Self-supervised pretraining offers a scalable approach with minimal labels, yet how to encode single-cell identity within tissue environments-and the extent of biological information such models can capture-remains an open question. Here, we introduce MAD (microenvironment-aware distillation), a pretraining strategy that learns cell-centric embeddings by jointly self-distilling the morphology view and the microenvironment view of the same indexed cell into a unified embedding space. Across diverse tissues and imaging modalities, MAD achieves state-of-the-art prediction performance on downstream tasks including cell subtyping, transcriptomic prediction, and bioinformatic inference. MAD even outperforms foundation models with a similar number of model parameters that have been trained on substantially larger datasets. These results demonstrate that MAD's dual-view joint self-distillation effectively captures the complexity and diversity of cells within tissues. Together, this establishes MAD as a general tool for representation learning in microscopy, enabling virtual spatial omics and biological insights from vast microscopy datasets.

cs.CV↗

arXiv2026

Extending the Cosmological Collider: New Scaling Regimes and Constraints from BOSS

Primordial non-Gaussianity generated by additional fields during inflation offers a compelling observational target. Heavy fields imprint characteristic oscillatory signals in non-Gaussian correlation functions of the inflaton, a process sometimes referred to as cosmological-collider physics. These distinct signatures are compelling windows into ultra-high-energy physics, but are often suppressed, making standard equilateral non-Gaussianity the most promising discovery channel in many scenarios. In this paper, we show that direct couplings between the inflaton and additional fields can lead to a wide variety of novel, observationally relevant signals which open new parameter regimes that simultaneously exhibit the characteristics of light and heavy fields. We identify these primordial signatures in the late-time observables of the large-scale structure of the Universe, where they most significantly modify the scale-dependent bias of the galaxy power spectrum to include an oscillatory modulation around a non-trivial power law. We explore the full range of parameters that phenomenologically arise in these models and study the sensitivity of current and future galaxy surveys, finding that this new class of primordial non-Gaussianity is particularly accessible in near-term surveys due to its oscillatory feature. Finally, we perform an analysis of existing data from the final release of the Baryon Oscillation Spectroscopic Survey (BOSS DR12). While we find no evidence for a signal, we demonstrate significant improvements in sensitivity over respective non-oscillatory scenarios and place the first constraints on this extended parameter space of oscillatory non-Gaussianity.

astro-ph.CO↗

arXiv2024

System- and Sample-agnostic Isotropic 3D Microscopy by Weakly Physics-informed, Domain-shift-resistant Axial Deblurring

Three-dimensional (3D) subcellular imaging is essential for biomedical research, but the diffraction limit of optical microscopy compromises axial resolution, hindering accurate 3D structural analysis. This challenge is particularly pronounced in label-free imaging of thick, heterogeneous tissues, where assumptions about data distribution (e.g. sparsity, label-specific distribution, and lateral-axial similarity) and system priors (e.g. independent and identically distributed (i.i.d.) noise and linear shift-invariant (LSI) point-spread functions (PSFs)) are often invalid. Here, we introduce SSAI-3D, a weakly physics-informed, domain-shift-resistant framework for robust isotropic 3D imaging. SSAI-3D enables robust axial deblurring by generating a PSF-flexible, noise-resilient, sample-informed training dataset and sparsely fine-tuning a large pre-trained blind deblurring network. SSAI-3D was applied to label-free nonlinear imaging of living organoids, freshly excised human endometrium tissue, and mouse whisker pads, and further validated in publicly available ground-truth-paired experimental datasets of 3D heterogeneous biological tissues with unknown blurring and noise across different microscopy systems.

physics.optics↗

arXiv2023

Light Fields during Inflation from BOSS and Future Galaxy Surveys

Primordial non-Gaussianity generated by additional fields present during inflation offers a compelling observational target for galaxy surveys. These fields are of significant theoretical interest since they offer a window into particle physics in the inflaton sector. They also violate the single-field consistency conditions and induce a scale-dependent bias in the galaxy power spectrum. In this paper, we explore this particular signal for light scalar fields and study the prospects for measuring it with galaxy surveys. We find that the sensitivities of current and future surveys are remarkably stable for different configurations, including between spectroscopic and photometric redshift measurements. This is even the case at non-zero masses where the signal is not obviously localized on large scales. For realistic galaxy number densities, we demonstrate that the redshift range and galaxy bias of the sample have the largest impact on the sensitivity in the power spectrum. These results additionally motivated us to explore the potentially enhanced sensitivity of Vera Rubin Observatory's LSST through multi-tracer analyses. Finally, we apply this understanding to current data from the last data release of the Baryon Oscillation Spectroscopic Survey (BOSS DR12) and place new constraints on light fields coupled to the inflaton.

astro-ph.CO↗

arXiv2023

Learned, uncertainty-driven adaptive acquisition for photon-efficient scanning microscopy

Scanning microscopy systems, such as confocal and multiphoton microscopy, are powerful imaging tools for probing deep into biological tissue. However, scanning systems have an inherent trade-off between acquisition time, field of view, phototoxicity, and image quality, often resulting in noisy measurements when fast, large field of view, and/or gentle imaging is needed. Deep learning could be used to denoise noisy microscopy measurements, but these algorithms can be prone to hallucination, which can be disastrous for medical and scientific applications. We propose a method to simultaneously denoise and predict pixel-wise uncertainty for scanning microscopy systems, improving algorithm trustworthiness and providing statistical guarantees for deep learning predictions. Furthermore, we propose to leverage this learned, pixel-wise uncertainty to drive an adaptive acquisition technique that rescans only the most uncertain regions of a sample, saving time and reducing the total light dose to the sample. We demonstrate our method on experimental confocal and multiphoton microscopy systems, showing that our uncertainty maps can pinpoint hallucinations in the deep learned predictions. Finally, with our adaptive acquisition technique, we demonstrate up to 16X reduction in acquisition time and total light dose while successfully recovering fine features in the sample and reducing hallucinations. We are the first to demonstrate distribution-free uncertainty quantification for a denoising task with real experimental data and the first to propose adaptive acquisition based on reconstruction uncertainty.

eess.IV↗

arXiv2023

Hypergraph Isomorphism Computation

The isomorphism problem is a fundamental problem in network analysis, which involves capturing both low-order and high-order structural information. In terms of extracting low-order structural information, graph isomorphism algorithms analyze the structural equivalence to reduce the solver space dimension, which demonstrates its power in many applications, such as protein design, chemical pathways, and community detection. For the more commonly occurring high-order relationships in real-life scenarios, the problem of hypergraph isomorphism, which effectively captures these high-order structural relationships, cannot be straightforwardly addressed using graph isomorphism methods. Besides, the existing hypergraph kernel methods may suffer from high memory consumption or inaccurate sub-structure identification, thus yielding sub-optimal performance. In this paper, to address the abovementioned problems, we first propose the hypergraph Weisfiler-Lehman test algorithm for the hypergraph isomorphism test problem by generalizing the Weisfiler-Lehman test algorithm from graphs to hypergraphs. Secondly, based on the presented algorithm, we propose a general hypergraph Weisfieler-Lehman kernel framework and implement two instances, which are Hypergraph Weisfeiler-Lehamn Subtree Kernel and Hypergraph Weisfeiler-Lehamn Hyperedge Kernel. In order to fulfill our research objectives, a comprehensive set of experiments was meticulously designed, including seven graph classification datasets and 12 hypergraph classification datasets. Results on hypergraph classification datasets show significant improvements compared to other typical kernel-based methods, which demonstrates the effectiveness of the proposed methods. In our evaluation, we found that our proposed methods outperform the second-best method in terms of runtime, running over 80 times faster when handling complex hypergraph structures.

cs.DS↗

arXiv2022

CLOSE: Curriculum Learning On the Sharing Extent Towards Better One-shot NAS

One-shot Neural Architecture Search (NAS) has been widely used to discover architectures due to its efficiency. However, previous studies reveal that one-shot performance estimations of architectures might not be well correlated with their performances in stand-alone training because of the excessive sharing of operation parameters (i.e., large sharing extent) between architectures. Thus, recent methods construct even more over-parameterized supernets to reduce the sharing extent. But these improved methods introduce a large number of extra parameters and thus cause an undesirable trade-off between the training costs and the ranking quality. To alleviate the above issues, we propose to apply Curriculum Learning On Sharing Extent (CLOSE) to train the supernet both efficiently and effectively. Specifically, we train the supernet with a large sharing extent (an easier curriculum) at the beginning and gradually decrease the sharing extent of the supernet (a harder curriculum). To support this training strategy, we design a novel supernet (CLOSENet) that decouples the parameters from operations to realize a flexible sharing scheme and adjustable sharing extent. Extensive experiments demonstrate that CLOSE can obtain a better ranking quality across different computational budget constraints than other one-shot supernets, and is able to discover superior architectures when combined with various search strategies. Code is available at https://github.com/walkerning/aw_nas.

cs.CV↗

arXiv2019

Probing Gravitational Lensing of the CMB with SDSS-IV Quasars

We study the cross-correlation between the Planck CMB lensing convergence map and the eBOSS quasar overdensity obtained from the Sloan Digital Sky Survey (SDSS) IV, in the redshift range $0.9 < z < 2.2$. We detect the CMB lensing convergence-quasar cross power spectrum at $5.4 σ$ significance. The cross power spectrum provides a quasar clustering bias measurement that is expected to be particularly robust against systematic effects. The redshift distribution of the quasar sample has a median redshift $z \approx 1.55$, and an effective redshift about $1.51$. The best fit bias of the quasar sample is $b_q = 2.43 \pm 0.45$, corresponding to a host halo mass of $\log_{10}\left( \frac{M}{h^{-1} M_\odot} \right) = 12.54^{+0.25}_{-0.36}$. This is broadly consistent with the previous literature on quasars with a similar redshift range and selection. Since our constraint on the bias comes from the cross-correlation between quasars and CMB lensing, we expect it to be robust to a wide range of possible systematic effects that may contaminate the auto correlation of quasars. We checked for a number of systematic effects from both CMB lensing and quasar overdensity, and found that all systematics are consistent with null within $2 σ$. The data is not sensitive to a possible scale dependence of the bias at present, but we expect that as the number of quasars increases (in future surveys such as DESI), it is likely that strong constraints on the scale dependence of the bias can be obtained.

astro-ph.CO↗