arXiv subjects

Zihui Wang

Publications and source records attributed to Zihui Wang.

14 recordsLinked to original sources

arXiv2025

HyperST: Hierarchical Hyperbolic Learning for Spatial Transcriptomics Prediction

Spatial Transcriptomics (ST) merges the benefits of pathology images and gene expression, linking molecular profiles with tissue structure to analyze spot-level function comprehensively. Predicting gene expression from histology images is a cost-effective alternative to expensive ST technologies. However, existing methods mainly focus on spot-level image-to-gene matching but fail to leverage the full hierarchical structure of ST data, especially on the gene expression side, leading to incomplete image-gene alignment. Moreover, a challenge arises from the inherent information asymmetry: gene expression profiles contain more molecular details that may lack salient visual correlates in histological images, demanding a sophisticated representation learning approach to bridge this modality gap. We propose HyperST, a framework for ST prediction that learns multi-level image-gene representations by modeling the data's inherent hierarchy within hyperbolic space, a natural geometric setting for such structures. First, we design a Multi-Level Representation Extractors to capture both spot-level and niche-level representations from each modality, providing context-aware information beyond individual spot-level image-gene pairs. Second, a Hierarchical Hyperbolic Alignment module is introduced to unify these representations, performing spatial alignment while hierarchically structuring image and gene embeddings. This alignment strategy enriches the image representations with molecular semantics, significantly improving cross-modal prediction. HyperST achieves state-of-the-art performance on four public datasets from different tissues, paving the way for more scalable and accurate spatial transcriptomics prediction.

cs.CV↗

arXiv2025

Federated Learning with Domain Shift Eraser

Federated learning (FL) is emerging as a promising technique for collaborative learning without local data leaving their devices. However, clients' data originating from diverse domains may degrade model performance due to domain shifts, preventing the model from learning consistent representation space. In this paper, we propose a novel FL framework, Federated Domain Shift Eraser (FDSE), to improve model performance by differently erasing each client's domain skew and enhancing their consensus. First, we formulate the model forward passing as an iterative deskewing process that extracts and then deskews features alternatively. This is efficiently achieved by decomposing each original layer in the neural network into a Domain-agnostic Feature Extractor (DFE) and a Domain-specific Skew Eraser (DSE). Then, a regularization term is applied to promise the effectiveness of feature deskewing by pulling local statistics of DSE's outputs close to the globally consistent ones. Finally, DFE modules are fairly aggregated and broadcast to all the clients to maximize their consensus, and DSE modules are personalized for each client via similarity-aware aggregation to erase their domain skew differently. Comprehensive experiments were conducted on three datasets to confirm the advantages of our method in terms of accuracy, efficiency, and generalizability.

cs.CV↗

arXiv2024

ST-Align: A Multimodal Foundation Model for Image-Gene Alignment in Spatial Transcriptomics

Spatial transcriptomics (ST) provides high-resolution pathological images and whole-transcriptomic expression profiles at individual spots across whole-slide scales. This setting makes it an ideal data source to develop multimodal foundation models. Although recent studies attempted to fine-tune visual encoders with trainable gene encoders based on spot-level, the absence of a wider slide perspective and spatial intrinsic relationships limits their ability to capture ST-specific insights effectively. Here, we introduce ST-Align, the first foundation model designed for ST that deeply aligns image-gene pairs by incorporating spatial context, effectively bridging pathological imaging with genomic features. We design a novel pretraining framework with a three-target alignment strategy for ST-Align, enabling (1) multi-scale alignment across image-gene pairs, capturing both spot- and niche-level contexts for a comprehensive perspective, and (2) cross-level alignment of multimodal insights, connecting localized cellular characteristics and broader tissue architecture. Additionally, ST-Align employs specialized encoders tailored to distinct ST contexts, followed by an Attention-Based Fusion Network (ABFN) for enhanced multimodal fusion, effectively merging domain-shared knowledge with ST-specific insights from both pathological and genomic data. We pre-trained ST-Align on 1.3 million spot-niche pairs and evaluated its performance through two downstream tasks across six datasets, demonstrating superior zero-shot and few-shot capabilities. ST-Align highlights the potential for reducing the cost of ST and providing valuable insights into the distinction of critical compositions within human tissue.

cs.CV↗

arXiv2024

Federated Graph Learning for Cross-Domain Recommendation

Cross-domain recommendation (CDR) offers a promising solution to the data sparsity problem by enabling knowledge transfer across source and target domains. However, many recent CDR models overlook crucial issues such as privacy as well as the risk of negative transfer (which negatively impact model performance), especially in multi-domain settings. To address these challenges, we propose FedGCDR, a novel federated graph learning framework that securely and effectively leverages positive knowledge from multiple source domains. First, we design a positive knowledge transfer module that ensures privacy during inter-domain knowledge transmission. This module employs differential privacy-based knowledge extraction combined with a feature mapping mechanism, transforming source domain embeddings from federated graph attention networks into reliable domain knowledge. Second, we design a knowledge activation module to filter out potential harmful or conflicting knowledge from source domains, addressing the issues of negative transfer. This module enhances target domain training by expanding the graph of the target domain to generate reliable domain attentions and fine-tunes the target model for improved negative knowledge filtering and more accurate predictions. We conduct extensive experiments on 16 popular domains of the Amazon dataset, demonstrating that FedGCDR significantly outperforms state-of-the-art methods.

cs.LG↗

arXiv2024

A new probe of dark matter-baryon interactions in compact stellar systems

We investigate the astrophysical consequences of an attractive long-range interaction between dark matter and baryonic matter. Our study highlights the role of this interaction in inducing dynamical friction between dark matter and stars, which can significantly influence the evolution of compact stellar systems. Using the star cluster in Eridanus II as a case study, we derive a new stringent upper bound on the interaction strength $\tilde{\alpha}\leq 314.5$ for the interaction range $\lambda = 1$ pc. This constraint is independent of the dark matter mass and can improve the existing model-independent limits on $\tilde{\alpha}$ by a few orders of magnitude. Furthermore, we observe that the constraint is insensitive to the mass of the stellar system and the dark matter density in the stellar system as long as the system is dark matter dominated. This new approach can be applied to many other stellar systems, and we obtain comparable constraints from compact stellar halos observed in ultrafaint dwarf galaxies.

hep-ph↗

arXiv2024

Selective Knowledge Sharing for Personalized Federated Learning Under Capacity Heterogeneity

Federated Learning (FL) stands to gain significant advantages from collaboratively training capacity-heterogeneous models, enabling the utilization of private data and computing power from low-capacity devices. However, the focus on personalizing capacity-heterogeneous models based on client-specific data has been limited, resulting in suboptimal local model utility, particularly for low-capacity clients. The heterogeneity in both data and device capacity poses two key challenges for model personalization: 1) accurately retaining necessary knowledge embedded within reduced submodels for each client, and 2) effectively sharing knowledge through aggregating size-varying parameters. To this end, we introduce Pa3dFL, a novel framework designed to enhance local model performance by decoupling and selectively sharing knowledge among capacity-heterogeneous models. First, we decompose each layer of the model into general and personal parameters. Then, we maintain uniform sizes for the general parameters across clients and aggregate them through direct averaging. Subsequently, we employ a hyper-network to generate size-varying personal parameters for clients using learnable embeddings. Finally, we facilitate the implicit aggregation of personal parameters by aggregating client embeddings through a self-attention module. We conducted extensive experiments on three datasets to evaluate the effectiveness of Pa3dFL. Our findings indicate that Pa3dFL consistently outperforms baseline methods across various heterogeneity settings. Moreover, Pa3dFL demonstrates competitive communication and computation efficiency compared to baseline approaches, highlighting its practicality and adaptability in adverse system conditions.

cs.LG↗

arXiv2024

FedSAC: Dynamic Submodel Allocation for Collaborative Fairness in Federated Learning

Collaborative fairness stands as an essential element in federated learning to encourage client participation by equitably distributing rewards based on individual contributions. Existing methods primarily focus on adjusting gradient allocations among clients to achieve collaborative fairness. However, they frequently overlook crucial factors such as maintaining consistency across local models and catering to the diverse requirements of high-contributing clients. This oversight inevitably decreases both fairness and model accuracy in practice. To address these issues, we propose FedSAC, a novel Federated learning framework with dynamic Submodel Allocation for Collaborative fairness, backed by a theoretical convergence guarantee. First, we present the concept of "bounded collaborative fairness (BCF)", which ensures fairness by tailoring rewards to individual clients based on their contributions. Second, to implement the BCF, we design a submodel allocation module with a theoretical guarantee of fairness. This module incentivizes high-contributing clients with high-performance submodels containing a diverse range of crucial neurons, thereby preserving consistency across local models. Third, we further develop a dynamic aggregation module to adaptively aggregate submodels, ensuring the equitable treatment of low-frequency neurons and consequently enhancing overall model accuracy. Extensive experiments conducted on three public benchmarks demonstrate that FedSAC outperforms all baseline methods in both fairness and model accuracy. We see this work as a significant step towards incentivizing broader client participation in federated learning. The source code is available at https://github.com/wangzihuixmu/FedSAC.

cs.LG↗

arXiv2023

Constraints on long-lived di-baryons and di-baryonic dark matter

A color-flavor-spin singlet state of six quarks $uuddss$ ($S$, or sexaquark) has been argued to be a potentially undiscovered deeply bound, long-lived hadron. Theoretical calculations of the $S$ mass have been made in the literature, widely varying from deeply-bound $\sim 1.2$ GeV to weakly-bound $\sim2.2$ GeV. Given the spread of the mass predictions, it is vital to derive observational constraints on the state as a function of the mass. The transition rates between $S$ and two baryons are governed by $\tilde{g}$, the effective Yukawa coupling between $S$ and two baryons with the same quantum numbers as $S$. In this paper, we place strong observational constraints on $\tilde{g}$, improving on various previous limits; additional limits assuming $S$ is dark matter are also presented.

hep-ph↗

arXiv2023

Constraining axion and compact dark matter with interstellar medium heating

Cold interstellar gas systems have been used to constrain dark matter (DM) models by the condition that the heating rate from DM must be lower than the astrophysical cooling rate of the gas. Following the methodology of Wadekar and Farrar (2021), we use the interstellar medium of a gas-rich dwarf galaxy, Leo T, and a Milky Way-environment gas cloud, G33.4-8.0 to constrain DM. Leo T is a particularly strong system as its gas can have the lowest cooling rate among all the objects in the late Universe (owing to the low volume density and metallicity of the gas). Milky Way clouds, in some cases, provide complementary limits as the DM-gas relative velocity in them is much larger than that in Leo T. We derive constraints on the following scenarios in which DM can heat the gas: $(i)$ interaction of axions with hydrogen atoms or free electrons in the gas, $(ii)$ deceleration of relic magnetically charged DM in gas plasma, $(iii)$ dynamical friction from compact DM, $(iv)$ hard sphere scattering of composite DM with gas. Our limits are complementary to DM direct detection searches. Detection of more gas-rich low-mass dwarfs like Leo T from upcoming 21cm and optical surveys can improve our bounds.

hep-ph↗

arXiv2022

Strong constraints on decay and annihilation of dark matter from heating of gas-rich dwarf galaxies

Gas-rich dwarf galaxies located outside the virial radius of their host are relatively pristine systems and have ultra-low gas cooling rates. This makes them very sensitive to heat injection by annihilation or decay of dark matter (DM). Such dwarfs are particularly sensitive to DM producing e$^\pm$ with energies 1$-$100 MeV or photons with energies 13.6 eV$-1$ keV, because these products can be efficiently thermalized in the neutral hydrogen gas of the dwarfs. Following the methodology of Wadekar and Farrar (2021), we require the rate of heat injection by DM to not exceed the ultra-low radiative cooling rate of gas in the Leo T dwarf galaxy. This gives model-independent bounds on $(i)$ the decay lifetime of DM to $e^\pm$ (photons) which are stronger than all the previous literature for $m_\mathrm{DM}\sim$ 1$-$10 MeV ($m_\mathrm{DM}\sim0.02-1$ keV), $(ii)$ annihilation of DM to $e^{\pm}$ comparable to constraints from CMB and X/$γ$-ray surveys. We also translate our bounds for the case of the following DM models: axion-like particles (ALPs), sterile neutrinos, excited DM states, higgs portal scalars and dark baryons. Observations of gas-rich low-mass dwarfs from upcoming 21cm and optical surveys can therefore be powerful probes of a multitude of models of DM.

hep-ph↗

arXiv2022

Dark Matter Particle in QCD

We report on the possibility that the Dark Matter particle is a stable, neutral, as-yet-undiscovered hadron in the standard model. The existence of a compact color-flavor-spin singlet sexaquark (S, uuddss) with mass ~2m_p, is compatible with current knowledge. The S interacts with baryons primarily via a Yukawa interaction of coupling strength alpha_SN, mediated by omega and phi vector mesons having mass ~1 GeV. If it exists, the S is a very attractive DM candidate. The relic abundance of S Dark Matter (SDM) is established when the Universe transitions from the quark-gluon plasma to the hadronic phase at ~150 MeV and is in remarkable agreement with the observed Omega_DM/Omega_b = 5.3+-0.1; this is a no-free-parameters result because the relevant parameters are known from QCD. Survival of this relic abundance to low temperature requires the breakup amplitude gtilde <~ 2 10^-6, comfortably compatible with theory expectations and observational bounds because the breakup amplitude is dynamically suppressed and many orders of magnitude smaller, as we show. The scattering cross section can differ by orders of magnitude from Born approximation, depending on alpha_SN, requiring reanalysis of observational limits. We use direct detection experiments and cosmological constraints to determine the allowed region of alpha_SN. For a range of allowed values, we predict exotic nuclear isotopes at a detectable level with mass offset ~2 amu. The most promising approaches for detecting the sexaquark in accelerator experiments are to search for a long-interaction-length neutral particle component in the central region of relativistic heavy ion collisions or using a beam-dump setup, and to search for evidence of missing particle production characterized by unbalanced baryon number and strangeness using Belle-II or possibly GLUEX at J-Lab.

hep-ph↗

arXiv2020

Subaru through a different lens: microlensing by extended dark matter structures

We investigate gravitational microlensing signals produced by a spatially extended object transiting in front of a finite-sized source star. The most interesting features arise for lens and source sizes comparable to the Einstein radius of the setup. Using this information, we obtain constraints from the Subaru-HSC survey of M31 on the dark matter populations of NFW subhalos and boson stars of asteroid to Earth masses. These lens profiles capture the qualitative behavior of a wide range of dark matter substructures. We find that deviations from constraints on point-like lenses (e.g. primordial black holes and MACHOs) become visible for lenses of radius 0.1 $R_\odot$ and larger, with the upper bound on lens masses weakening with increasing lens size.

astro-ph.CO↗

arXiv2015

Formation of Bright Solitons from Wave Packets with Repulsive Nonlinearity

Formation of bright envelope solitons from wave packets with a repulsive nonlinearity was observed for the first time. The experiments used surface spin-wave packets in magnetic yttrium iron garnet (YIG) thin film strips. When the wave packets are narrow and have low power, they undergo self-broadening during the propagation. When the wave packets are relatively wide or their power is relatively high, they can experience self-narrowing or even evolve into bright solitons. The experimental results were reproduced by numerical simulations based on a modified nonlinear Schrödinger equation model.

cond-mat.mes-hall↗

arXiv2011

Chaotic Spin-Wave Solitons in Magnetic Film Feedback Rings

Chaotic spin-wave solitons in magnetic film active feedback rings were observed for the first time. At some ring gain level, one observes the self-generation of a single spin-wave soliton pulse in the ring. When the pulse circulates in the ring, its amplitude varies chaotically with time. Numerical simulations based on a gain-loss nonlinear Schrödinger equation reproduce the observed responses.

cond-mat.mes-hall↗