arXiv subjects

Zhen Zhang

Publications and source records attributed to Zhen Zhang.

At least 253 records · Page 14Linked to original sources

arXiv2025

Analytic DAG Constraints for Differentiable DAG Learning

Recovering the underlying Directed Acyclic Graph (DAG) structures from observational data presents a formidable challenge, partly due to the combinatorial nature of the DAG-constrained optimization problem. Recently, researchers have identified gradient vanishing as one of the primary obstacles in differentiable DAG learning and have proposed several DAG constraints to mitigate this issue. By developing the necessary theory to establish a connection between analytic functions and DAG constraints, we demonstrate that analytic functions from the set $\{f(x) = c_0 + \sum_{i=1}^{\infty}c_ix^i | \forall i > 0, c_i > 0; r = \lim_{i\rightarrow \infty}c_{i}/c_{i+1} > 0\}$ can be employed to formulate effective DAG constraints. Furthermore, we establish that this set of functions is closed under several functional operators, including differentiation, summation, and multiplication. Consequently, these operators can be leveraged to create novel DAG constraints based on existing ones. Using these properties, we design a series of DAG constraints and develop an efficient algorithm to evaluate them. Experiments in various settings demonstrate that our DAG constraints outperform previous state-of-the-art comparators. Our implementation is available at https://github.com/zzhang1987/AnalyticDAGLearning.

cs.LG↗

arXiv2025

Collaborative AI Enhances Image Understanding in Materials Science

The Copilot for Real-world Experimental Scientist (CRESt) system empowers researchers to control autonomous laboratories through conversational AI, providing a seamless interface for managing complex experimental workflows. We have enhanced CRESt by integrating a multi-agent collaboration mechanism that utilizes the complementary strengths of the ChatGPT and Gemini models for precise image analysis in materials science. This innovative approach significantly improves the accuracy of experimental outcomes by fostering structured debates between the AI models, which enhances decision-making processes in materials phase analysis. Additionally, to evaluate the generalizability of this approach, we tested it on a quantitative task of counting particles. Here, the collaboration between the AI models also led to improved results, demonstrating the versatility and robustness of this method. By harnessing this dual-AI framework, this approach stands as a pioneering method for enhancing experimental accuracy and efficiency in materials research, with applications extending beyond CRESt to broader scientific experimentation and analysis.

cs.AI↗

arXiv2025

PyGDA: A Python Library for Graph Domain Adaptation

Graph domain adaptation has emerged as a promising approach to facilitate knowledge transfer across different domains. Recently, numerous models have been proposed to enhance their generalization capabilities in this field. However, there is still no unified library that brings together existing techniques and simplifies their implementation. To fill this gap, we introduce PyGDA, an open-source Python library tailored for graph domain adaptation. As the first comprehensive library in this area, PyGDA covers more than 20 widely used graph domain adaptation methods together with different types of graph datasets. Specifically, PyGDA offers modular components, enabling users to seamlessly build custom models with a variety of commonly used utility functions. To handle large-scale graphs, PyGDA includes support for features such as sampling and mini-batch processing, ensuring efficient computation. In addition, PyGDA also includes comprehensive performance benchmarks and well-documented user-friendly API for both researchers and practitioners. To foster convenient accessibility, PyGDA is released under the MIT license at https://github.com/pygda-team/pygda, and the API documentation is https://pygda.readthedocs.io/en/stable/.

cs.LG↗

arXiv2025

Adaptive$^2$: Adaptive Domain Mining for Fine-grained Domain Adaptation Modeling

Advertising systems often face the multi-domain challenge, where data distributions vary significantly across scenarios. Existing domain adaptation methods primarily focus on building domain-adaptive neural networks but often rely on hand-crafted domain information, e.g., advertising placement, which may be sub-optimal. We think that fine-grained "domain" patterns exist that are difficult to hand-craft in online advertisement. Thus, we propose Adaptive$^2$, a novel framework that first learns domains adaptively using a domain mining module by self-supervision and then employs a shared&specific network to model shared and conflicting information. As a practice, we use VQ-VAE as the domain mining module and conduct extensive experiments on public benchmarks. Results show that traditional domain adaptation methods with hand-crafted domains perform no better than single-domain models under fair FLOPS conditions, highlighting the importance of domain definition. In contrast, Adaptive$^2$ outperforms existing approaches, emphasizing the effectiveness of our method and the significance of domain mining. We also deployed Adaptive$^2$ in the live streaming scenario of Kuaishou Advertising System, demonstrating its commercial value and potential for automatic domain identification. To the best of our knowledge, Adaptive$^2$ is the first approach to automatically learn both domain identification and adaptation in online advertising, opening new research directions for this area.

cs.LG↗

arXiv2025

Evolution of the Three Spectral Components in the Prompt Emission of GRB 240825A

The prompt emission of Gamma-Ray Bursts (GRBs) could be composed of different spectral components, such as a dominant non-thermal Band component in the keV-MeV range, a subdominant quasi-thermal component, and an additional hard non-thermal component extending into the GeV range. The existence and evolutionary behaviors of these components could place strong implication on physical models, such as ejecta composition and dissipation processes. Although numerous GRBs have been found to exhibit one or two spectral components, reports of GRBs containing all three components remain rare. In this letter, based on the joint observations of GRB 240825A from multiple gamma-ray telescopes, we conduct a comprehensive temporal and spectral analysis to identify the presence and evolution of all three components. The bulk Lorentz factor of this bright and relatively short-duration burst is independently calculated using the thermal and hard non-thermal components, supporting a jet penetration scenario. The multi-segment broken powerlaw feature observed in the flux light curves suggests the presence of an early afterglow in the keV-MeV band and hints at a possible two-jet structure. Furthermore, the observed transition from positive to negative on the spectral lag can be interpreted as an independent evolution of the soft and hard components, leading to misalignment in the cross-correlation function (CCF) analysis of pulses.

astro-ph.HE↗

arXiv2025

The GECAM Ground Search System for Gamma-ray Transients

In the era of time-domain, multi-messenger astronomy, the detection of transient events on the high-energy electromagnetic sky has become more important than ever. The Gravitational wave high-energy Electromagnetic Counterpart All-sky Monitor (GECAM) is a dedicated mission to monitor gamma-ray transients, launched in December, 2020. A real-time on-board trigger and location software, using the traditional signal-to-noise ratio (SNR) method for blind search, is constrained to relatively bright signals due to the limitations in on-board computing resources and the need for real-time search. In this work, we developed a ground-based pipeline for GECAM to search for various transients, especially for weak bursts missed by on-board software. This pipeline includes both automatic and manual mode, offering options for blind search and targeted search. The targeted search is specifically designed to search for interesting weak bursts, such as gravitational wave-associated gamma-ray bursts (GRBs). From the ground search of the data in the first year, GECAM has been triggered by 54 GRBs and other transients, including soft gamma-ray repeaters, X-ray binaries, solar flares, terrestrial gamma-ray flashes. We report the properties of each type of triggers,such as trigger time and light curves. With this search pipeline and assuming a soft Band spectrum, the GRB detection sensitivity of GECAM is increased to about 1.1E-08 erg cm-2 s-1 (10 keV - 1000 keV, burst duration of 20 s). These results demonstrate that the GECAM ground search system (both blind search and targeted search) is a versatile pipeline to recover true astrophysical signals which were too weak to be found in the on-board search.

astro-ph.HE↗

arXiv2025

Online Pseudo-average Shifting Attention(PASA) for Robust Low-precision LLM Inference: Algorithms and Numerical Analysis

Attention calculation is extremely time-consuming for long-sequence inference tasks, such as text or image/video generation, in large models. To accelerate this process, we developed a low-precision, mathematically-equivalent algorithm called PASA, based on Flash Attention. PASA introduces two novel techniques: online pseudo-average shifting and global recovering. These techniques enable the use of half-precision computation throughout the Flash Attention process without incurring overflow instability or unacceptable numerical accuracy loss. This algorithm enhances performance on memory-restricted AI hardware architectures, such as the Ascend Neural-network Processing Unit(NPU), by reducing data movement and increasing computational FLOPs. The algorithm is validated using both designed random benchmarks and real large models. We find that the large bias and amplitude of attention input data are critical factors contributing to numerical overflow ($>65504$ for half precision) in two different categories of large models (Qwen2-7B language models and Stable-Video-Diffusion multi-modal models). Specifically, overflow arises due to the large bias in the sequence dimension and the resonance mechanism between the query and key in the head dimension of the Stable-Video-Diffusion models. The resonance mechanism is defined as phase coincidence or 180-degree phase shift between query and key matrices. It will remarkably amplify the element values of attention score matrix. This issue also applies to the Qwen models. Additionally, numerical accuracy is assessed through root mean square error (RMSE) and by comparing the final generated texts and videos to those produced using high-precision attention.

cs.LG↗

arXiv2025

A New Paradigm for Reconfigurable Intelligent Surface Design: Multi-port Network Method

As a novel approach to flexibly adjust the wireless environment, reconfigurable intelligent surfaces (RIS) have shown significant application potential across various domains, including wireless communication, radar detection, and the Internet of Things. Currently, mainstream design methods for reconfigurable intelligent surfaces face inherent limitations. For instance, while the full-wave electromagnetic (EM) simulation method offers strong universality, it suffers from low efficiency. Machine learning-based methods can effectively reduce design time but are heavily dependent on full-wave EM simulations. Although the design methods based on the equivalent circuit can lessen the reliance on full-wave EM simulations, they still struggle with insufficient model accuracy when dealing with complex element structures. In recent years, a new multi-port network method has been introduced to RIS design. This method has significantly enhanced the accuracy of modeling complex structures. It reduces the dependency on full-wave EM simulations and substantially shortens the design time. This work provides a detailed exploration of the RIS element design strategy based on multi-port network and discusses future development trends in this field.

physics.app-ph↗

arXiv2025

MaZO: Masked Zeroth-Order Optimization for Multi-Task Fine-Tuning of Large Language Models

Large language models have demonstrated exceptional capabilities across diverse tasks, but their fine-tuning demands significant memory, posing challenges for resource-constrained environments. Zeroth-order (ZO) optimization provides a memory-efficient alternative by eliminating the need for backpropagation. However, ZO optimization suffers from high gradient variance, and prior research has largely focused on single-task learning, leaving its application to multi-task learning unexplored. Multi-task learning is crucial for leveraging shared knowledge across tasks to improve generalization, yet it introduces unique challenges under ZO settings, such as amplified gradient variance and collinearity. In this paper, we present MaZO, the first framework specifically designed for multi-task LLM fine-tuning under ZO optimization. MaZO tackles these challenges at the parameter level through two key innovations: a weight importance metric to identify critical parameters and a multi-task weight update mask to selectively update these parameters, reducing the dimensionality of the parameter space and mitigating task conflicts. Experiments demonstrate that MaZO achieves state-of-the-art performance, surpassing even multi-task learning methods designed for first-order optimization.

cs.LG↗

arXiv2025

PSR J1231-1411 revisited: Pulse Profile Analysis of X-ray Observation

One of the primary goals of Neutron Star Interior Composition Explorer (NICER)-like X-ray missions is to impose stringent constraints on the neutron star equation of state by precisely measuring their masses and radii. NICER has recently expanded the dataset of inferred mass-radius relations for neutron stars, including four rotation-powered millisecond pulsars PSR J0030+0451, PSR J0740+6620, PSR J0437-4715, and PSR J1231-1411. In this work, the mass-radius relation and X-ray emitting region properties of PSR J1231-1411 are inferred with an independent pulse profile modeling based on the spherical star Schwarzschild-spacetime and Doppler approximation. With one single-temperature elongated hot spot and one single-temperature crescent hot spot, the inferred gravitational mass is $M = 1.12 \pm 0.07 M_{\odot}$ and the inferred equatorial radius is $R_{eq} = 9.91_{-0.86}^{+0.88}$ km (68% credible intervals). It provides an alternative geometry configuration of the X-ray emitting region for PSR J1231-1411 to sufficiently explain the observation data of NICER and XMM-Newton. The inferred radius is smaller than that derived by \citet{salmi2024nicer} ($M = 1.04_{-0.03}^{+0.05} M_{\odot}$, $R_{eq} = 12.6 \pm 0.3$ km), and the inferred mass is slightly higher in this work. The inferred geometry configurations of the X-ray emitting region in both works are non-antipodal, which is not consistent with a centered dipole magnetic field and suggests a complex magnetic field structure.

astro-ph.HE↗

arXiv2025

Integrating Neural Operators with Diffusion Models Improves Spectral Representation in Turbulence Modeling

We integrate neural operators with diffusion models to address the spectral limitations of neural operators in surrogate modeling of turbulent flows. While neural operators offer computational efficiency, they exhibit deficiencies in capturing high-frequency flow dynamics, resulting in overly smooth approximations. To overcome this, we condition diffusion models on neural operators to enhance the resolution of turbulent structures. Our approach is validated for different neural operators on diverse datasets, including a high Reynolds number jet flow simulation and experimental Schlieren velocimetry. The proposed method significantly improves the alignment of predicted energy spectra with true distributions compared to neural operators alone. This enables the diffusion models to stabilize longer forecasts through diffusion-corrected autoregressive rollouts, as we demonstrate in this work. Additionally, proper orthogonal decomposition analysis demonstrates enhanced spectral fidelity in space-time. This work establishes a new paradigm for combining generative models with neural operators to advance surrogate modeling of turbulent systems, and it can be used in other scientific applications that involve microstructure and high-frequency content. See our project page: vivekoommen.github.io/NO_DM

cs.LG↗

arXiv2025

Spectrotemporal shaping of attosecond x-ray pulses with a fresh-slice free-electron laser

We propose a scheme allowing coherent shaping, i.e., controlling both the amplitude and phase, of attosecond x-ray pulses at free-electron lasers. We show that by seeding an FEL with a short coherent seed that overfills the amplification bandwidth, one can shape the Wigner function of the pulse by controlling the undulator taper profile. The examples of controllable pulse pairs and trains, as well as isolated spectrotemporally shaped pulses with very broad bandwidths are examined in detail. Existing attosecond XFELs can achieve these experimental conditions in a two-stage cascade, in which the seed is generated by a short current spike in an electron bunch and shaped in an unspoiled region within the same bunch. We experimentally demonstrate the production and control of phase-stable pulse trains using this method at the Linac Coherent Light Source II.

physics.acc-ph↗

arXiv2025

Aggregate to Adapt: Node-Centric Aggregation for Multi-Source-Free Graph Domain Adaptation

Unsupervised graph domain adaptation (UGDA) focuses on transferring knowledge from labeled source graph to unlabeled target graph under domain discrepancies. Most existing UGDA methods are designed to adapt information from a single source domain, which cannot effectively exploit the complementary knowledge from multiple source domains. Furthermore, their assumptions that the labeled source graphs are accessible throughout the training procedure might not be practical due to privacy, regulation, and storage concerns. In this paper, we investigate multi-source-free unsupervised graph domain adaptation, i.e., adapting knowledge from multiple source domains to an unlabeled target domain without utilizing labeled source graphs but relying solely on source pre-trained models. Unlike previous multi-source domain adaptation approaches that aggregate predictions at model level, we introduce a novel model named GraphATA which conducts adaptation at node granularity. Specifically, we parameterize each node with its own graph convolutional matrix by automatically aggregating weight matrices from multiple source models according to its local context, thus realizing dynamic adaptation over graph structured data. We also demonstrate the capability of GraphATA to generalize to both model-centric and layer-centric methods. Comprehensive experiments on various public datasets show that our GraphATA can consistently surpass recent state-of-the-art baselines with different gains.

cs.LG↗

arXiv2025

The electronic structure, crystal fields, and magnetic anisotropy in RECo$_5$ magnets

The current progress in describing rare-earth-based magnets' electronic structure and magnetic properties is discussed. We use several currently popular electronic structure methods to show the typical values of critical parameters that define the physics of RECo$_5$ (RE = rare earth atom) materials. The magnetic moments and anisotropy of 4\textit{f} atoms are obtained using several approaches, including anisotropic 4\textit{f}-charge density-constrained DFT and DFT+HI methods. We also suggest the introduction of 'penalty' functional for obtaining correct variational total energy in the traditional Hund's rule-constrained DFT-based techniques. The applicability and future extensions are discussed. The proposed combination of methods is potentially suitable for high-throughput computational searches of new rare-earth-containing magnetic materials.

cond-mat.mtrl-sci↗

arXiv2025

Physics-Informed Machine Learning for Efficient Reconfigurable Intelligent Surface Design

Reconfigurable intelligent surface (RIS) is a two-dimensional periodic structure integrated with a large number of reflective elements, which can manipulate electromagnetic waves in a digital way, offering great potentials for wireless communication and radar detection applications. However, conventional RIS designs highly rely on extensive full-wave EM simulations that are extremely time-consuming. To address this challenge, we propose a machine-learning-assisted approach for efficient RIS design. An accurate and fast model to predict the reflection coefficient of RIS element is developed by combining a multi-layer perceptron neural network (MLP) and a dual-port network, which can significantly reduce tedious EM simulations in the network training. A RIS has been practically designed based on the proposed method. To verify the proposed method, the RIS has also been fabricated and measured. The experimental results are in good agreement with the simulation results, which validates the efficacy of the proposed method in RIS design.

cs.LG↗

arXiv2025

Unsupervised Query Routing for Retrieval Augmented Generation

Query routing for retrieval-augmented generation aims to assign an input query to the most suitable search engine. Existing works rely heavily on supervised datasets that require extensive manual annotation, resulting in high costs and limited scalability, as well as poor generalization to out-of-distribution scenarios. To address these challenges, we introduce a novel unsupervised method that constructs the "upper-bound" response to evaluate the quality of retrieval-augmented responses. This evaluation enables the decision of the most suitable search engine for a given query. By eliminating manual annotations, our approach can automatically process large-scale real user queries and create training data. We conduct extensive experiments across five datasets, demonstrating that our method significantly enhances scalability and generalization capabilities.

cs.IR↗

arXiv2025

High-Throughput Studies of Novel Magnetic Materials in Borides

Borides are a versatile material family with various properties for valuable applications. Conventional magnetism, such as ferromagnetism and antiferromagnetism in borides, have been extensively studied. However, research on unconventional magnetism in borides where quantum effects are dominant is scarce. Here, we implement a high-throughput workflow combining first-principles calculations, materials prediction, and magnetic properties calculations to discover novel magnetism and magnetic materials in borides. Successfully applying the workflow, we report three families of novel magnetic borides, including two families of borides exhibiting quantum magnetism. One is a family of dimerized quantum magnets among YCrB$_4$-type borides, which provides a rare platform for studying the spin-gap quantum critical point. The other is a family of altermagnets among FeMo$_2$B$_2$-type borides, extending the magnetic orderings exhibited by borides beyond conventional ferromagnetism and antiferromagnetism. We also predict a family of magnetic laminate transition metal borides, known as the MAB phases, in the AlFe$_2$B$_2$-type family, which provide pure-phase or alloying candidates for studying magnetocaloric materials and the associated magnetic transitions. The workflow is expected to be used in further studies of novel magnetism and magnetic materials.

cond-mat.mtrl-sci↗

arXiv2025

Highway Graph to Accelerate Reinforcement Learning

Reinforcement Learning (RL) algorithms often struggle with low training efficiency. A common approach to address this challenge is integrating model-based planning algorithms, such as Monte Carlo Tree Search (MCTS) or Value Iteration (VI), into the environmental model. However, VI requires iterating over a large tensor which updates the value of the preceding state based on the succeeding state through value propagation, resulting in computationally intensive operations. To enhance the RL training efficiency, we propose improving the efficiency of the value learning process. In deterministic environments with discrete state and action spaces, we observe that on the sampled empirical state-transition graph, a non-branching sequence of transitions-termed a highway-can take the agent to another state without deviation through intermediate states. On these non-branching highways, the value-updating process can be streamlined into a single-step operation, eliminating the need for step-by-step updates. Building on this observation, we introduce the highway graph to model state transitions. The highway graph compresses the transition model into a compact representation, where edges can encapsulate multiple state transitions, enabling value propagation across multiple time steps in a single iteration. By integrating the highway graph into RL, the training process is significantly accelerated, particularly in the early stages of training. Experiments across four categories of environments demonstrate that our method learns significantly faster than established and state-of-the-art RL algorithms (often by a factor of 10 to 150) while maintaining equal or superior expected returns. Furthermore, a deep neural network-based agent trained using the highway graph exhibits improved generalization capabilities and reduced storage costs. Code is publicly available at https://github.com/coodest/highwayRL.

cs.LG↗