arXiv subjects

Wenhui Fan

Publications and source records attributed to Wenhui Fan.

11 recordsLinked to original sources

arXiv2026

Continual-Learning Physics-Informed Neural Networks for Parameterized Partial Differential Equations

Physics-informed neural networks (PINNs) incorporate governing equations into neural-network training and can approximate PDE solutions without requiring large observational datasets. Parameterized PINNs (ParamPINNs) further take physical parameters as inputs, allowing a single model to represent a family of PDE solutions over a parameter domain. Existing ParamPINNs, however, still face inefficient training, uneven accuracy across parameters, and overfitting to a limited set of sampled parameter tasks, which can impair generalization to unsampled parameters. To address these issues, we propose a continual-learning physics-informed neural network (CL-PINN), which treats PDE instances at different parameter values as related tasks and learns them sequentially. CL-PINN combines Bayesian-optimization-based active parameter selection, task-wise dynamic loss weighting, sparse physics-constrained replay, and an optional parameter subnetwork to improve task allocation and knowledge retention under bounded active-task capacity. It requires no observational data and is designed to solve parameterized PDEs over relatively broad parameter domains under limited computational resources. Multi-seed evaluations on five benchmarks, including one continuous function and four parameterized PDEs, show that Bayesian selection substantially reduces objective-loss queries relative to grid-greedy search, while sparse replay mitigates forgetting of earlier tasks. Under the prescribed within-case resource protocols, CL-PINN generally provides higher and more balanced solution accuracy than fixed-sampling and grid-greedy baselines. CL-PINN offers a practical route toward learning PDE solutions that generalize across physical parameters and has the potential to support reusable physics-informed surrogates for large-scale engineering parameter studies.

cs.LG↗

arXiv2026

Curriculum Learning of Physics-Informed Neural Networks based on Spatial Correlation

Physics-Informed Neural Networks (PINNs) combine deep learning with physical constraints for solving partial differential equations (PDEs), and are widely applied in fluid mechanics, heat transfer, and solid mechanics. However, PINN training still suffers from high-dimensional non-convex loss landscapes, imbalanced multiobjective constraints, and ineffective information propagation. Existing curriculum learning and causality-guided strategies improve training stability, but mainly focus on temporal or parametric progression, lacking explicit treatment of spatial information propagation and inter-region consistency. Moreover, they are not directly applicable to boundary value problems (BVPs) with strong spatial coupling. To address this issue, we propose a spatially correlated curriculum learning framework for PINNs. To the best of our knowledge, this is the first work to address PINN training difficulties from the perspective of spatial coupling among subregions. First, spatial causal weights guide information from near-boundary regions inward, reducing optimization failures and spurious convergence. Second, a low-frequency information bridge enforces pseudo-label-based consistency across spatially separated regions, suppressing global low-frequency drift. Third, a region-adaptive reweighting strategy adjusts subregion losses to reduce local residuals and recover high-frequency details. Experiments on PDE benchmarks show that, under comparable computational cost, the proposed method alleviates training failures and improves solution accuracy. The code is available at https://github.com/pigofmomo/CurriculumLearningPINN.

cs.LG↗

arXiv2024

Heterogeneous Multi-Agent Reinforcement Learning for Zero-Shot Scalable Collaboration

The emergence of multi-agent reinforcement learning (MARL) is significantly transforming various fields like autonomous vehicle networks. However, real-world multi-agent systems typically contain multiple roles, and the scale of these systems dynamically fluctuates. Consequently, in order to achieve zero-shot scalable collaboration, it is essential that strategies for different roles can be updated flexibly according to the scales, which is still a challenge for current MARL frameworks. To address this, we propose a novel MARL framework named Scalable and Heterogeneous Proximal Policy Optimization (SHPPO), integrating heterogeneity into parameter-shared PPO-based MARL networks. We first leverage a latent network to learn strategy patterns for each agent adaptively. Second, we introduce a heterogeneous layer to be inserted into decision-making networks, whose parameters are specifically generated by the learned latent variables. Our approach is scalable as all the parameters are shared except for the heterogeneous layer, and gains both inter-individual and temporal heterogeneity, allowing SHPPO to adapt effectively to varying scales. SHPPO exhibits superior performance in classic MARL environments like Starcraft Multi-Agent Challenge (SMAC) and Google Research Football (GRF), showcasing enhanced zero-shot scalability, and offering insights into the learned latent variables' impact on team performance by visualization.

cs.LG↗

arXiv2024

Embodied LLM Agents Learn to Cooperate in Organized Teams

Large Language Models (LLMs) have emerged as integral tools for reasoning, planning, and decision-making, drawing upon their extensive world knowledge and proficiency in language-related tasks. LLMs thus hold tremendous potential for natural language interaction within multi-agent systems to foster cooperation. However, LLM agents tend to over-report and comply with any instruction, which may result in information redundancy and confusion in multi-agent cooperation. Inspired by human organizations, this paper introduces a framework that imposes prompt-based organization structures on LLM agents to mitigate these problems. Through a series of experiments with embodied LLM agents and human-agent collaboration, our results highlight the impact of designated leadership on team efficiency, shedding light on the leadership qualities displayed by LLM agents and their spontaneous cooperative behaviors. Further, we harness the potential of LLMs to propose enhanced organizational prompts, via a Criticize-Reflect process, resulting in novel organization structures that reduce communication costs and enhance team efficiency.

cs.AI↗

arXiv2023

Scalable Communication for Multi-Agent Reinforcement Learning via Transformer-Based Email Mechanism

Communication can impressively improve cooperation in multi-agent reinforcement learning (MARL), especially for partially-observed tasks. However, existing works either broadcast the messages leading to information redundancy, or learn targeted communication by modeling all the other agents as targets, which is not scalable when the number of agents varies. In this work, to tackle the scalability problem of MARL communication for partially-observed tasks, we propose a novel framework Transformer-based Email Mechanism (TEM). The agents adopt local communication to send messages only to the ones that can be observed without modeling all the agents. Inspired by human cooperation with email forwarding, we design message chains to forward information to cooperate with the agents outside the observation range. We introduce Transformer to encode and decode the message chain to choose the next receiver selectively. Empirically, TEM outperforms the baselines on multiple cooperative MARL benchmarks. When the number of agents varies, TEM maintains superior performance without further training.

cs.MA↗

arXiv2022

Anomalous quasiparticles in the zone center electron pocket of the kagom\'e ferromagnet Fe3Sn2

One material containing kagome bilayers and featuring both exceptional magnetism and electron transport is the ferromagnetic metal Fe3Sn2. Notwithstanding the widespread interest in Fe3Sn2, crystal twinning, difficulties in distinguishing surface from bulk states, and a large unit cell have until now prevented the synchrotron-based spectroscopic observation of sharply resolved quasiparticles near the Fermi surface which could be responsible for the anomalous properties appearing at low temperatures for the material. Here we report microfocused laser-based angle-resolved photoemission spectroscopy (micro-ARPES), which offers the first look at such quasiparticles. The high spatial resolution allows individual crystal twin domains to be examined in isolation, resulting in the discovery of three-fold symmetric electron pockets at the Brillouin zone (BZ) center, not predicted by early tight-binding descriptions but in agreement with density functional theory (DFT) calculations, which also feature Weyl nodes. The quasiparticles in these pockets have remarkably long mean free paths, and their Fermi surface area is consistent with reported quantum oscillations. At the same time, though, the best-defined Fermi surface is reduced at low temperature, and the quasiparticles generally are marginal in the sense that their wavelength uncertainty is of order the deviation of the quasiparticle wavelength from the Fermi vector. We attribute these manifestations of strong electron-electron interactions to a flat band predicted by our DFT to lie just above the dispersive bands seen in this experiment. Thus, beyond demonstrating the impact of twin averaging for ARPES measurements of band structures, our experiments reveal many-body physics unaccounted for by current theories of metallic kagome ferromagnets.

cond-mat.str-el↗

arXiv2021

Discovery of two families of VSb-based compounds with V-kagome lattice

We report the structure and physical properties of two newly-discovered compounds AV8Sb12 and AV6Sb6 (A = Cs, Rb), which have C2 (space group: Cmmm) and C3 (space group: R-3m) symmetry, respectively. The basic V-kagome unit is present in both compounds, but stacking differently. A V2Sb2 layer is sandwiched between two V3Sb5 layers in AV8Sb12, altering the V-kagome lattice and lowering the symmetry of kagome layer from hexagonal to orthorhombic. In AV6Sb6, the building block is a more complex slab made up of two half-V3Sb5 layers that are intercalated by Cs cations along the c-axis. Transport property measurements demonstrate that both compounds are nonmagnetic metals, with carrier concentrations at around 1021cm-3. No superconductivity has been observed in CsV8Sb12 above 0.3 K under in-situ pressure up to 46 GPa. In contrast to CsV3Sb5, theoretical calculations and angle-resolved photoemission spectroscopy (ARPES) reveal a quasi-two-dimensional electronic structure in CsV8Sb12 with C2 symmetry and no van Hove singularities near the Fermi level. Our findings will stimulate more research into V-based kagome quantum materials.

cond-mat.supr-con↗

arXiv2021

Charge instability of topological Fermi arcs in chiral crystal CoSi

Topological boundary states, emerged at the spatial boundary between topological non-trivial and trivial phases, are usually gapless, or commonly referred as metallic states. For example, the surface state of a topological insulator is a gapless Dirac state. These metallic topological boundary states are typically well described by non-interacting fermions. However, the behavior of topological boundary states with significant electron-electron interactions, which could turn the gapless boundary states into gapped ordered states, e.g., density wave states or superconducting states, is of great interest theoretically, but is still lacking evidence experimentally. Here, we report the observation of incommensurable charge density wave (CDW) formed on the topological boundary states driven by the electron-electron interactions on the (001) surface of CoSi. The wavevector of CDW varies as the temperature changes, which coincides with the evolution of topological surface Fermi arcs with temperature. The orientation of CDW phase is determined by the chirality of the Fermi arcs, which indicates direct association between CDW and Fermi arcs. Our finding will stimulate the search of more interactions-driven ordered states, such as superconductivity and magnetism, on the boundaries of topological materials.

cond-mat.str-el↗

arXiv2021

Discovery of \^C$_2$ rotation anomaly in topological crystalline insulator SrPb

Topological crystalline insulators (TCIs) are insulating electronic states with nontrivial topology protected by crystalline symmetries. Recently, theory has proposed new classes of TCIs protected by rotation symmetries \^C$_n$, which have surface rotation anomaly evading the fermion doubling theorem, i.e. n instead of 2n Dirac cones on the surface preserving the rotation symmetry. Here, we report the first realization of the \^C$_2$ rotation anomaly in a binary compound SrPb. Our first-principles calculations reveal two massless Dirac fermions protected by the combination of time-reversal symmetry \^T and \^C$_{2y}$ on the (010) surface. Using angle-resolved photoemission spectroscopy, we identify two Dirac surface states inside the bulk band gap of SrPb, confirming the \^C$_2$ rotation anomaly in the new classes of TCIs. The findings enrich the classification of topological phases, which pave the way for exploring exotic behaviour of the new classes of TCIs.

cond-mat.mtrl-sci↗

arXiv2020

Multiple mobile excitons manifested as sidebands in quasi-one-dimensional metallic TaSe3

Charge neutrality and their expected itinerant nature makes excitons potential transmitters of information. However, exciton mobility remains inaccessible to traditional optical experiments that only create and detect excitons with negligible momentum. Here, using angle-resolved photoemission spectroscopy, we detect dispersing excitons in the quasi-one-dimensional metallic trichalcogenide, TaSe3. The low density of conduction electrons and the low dimensionality in TaSe3 combined with a polaronic renormalization of the conduction band and the poorly screened interaction between these polarons and photo-induced valence holes leads to various excitonic bound states that we interpret as intrachain and interchain excitons, and possibly trions. The thresholds for the formation of a photo-hole together with an exciton appear as side valence bands with dispersions nearly parallel to the main valence band, but shifted to lower excitation energies. The energy separation between side and main valence bands can be controlled by surface doping, enabling the tuning of certain exciton properties.

cond-mat.str-el↗

arXiv2019

Magnetic topological insulator MnBi6Te10 with zero-field ferromagnetic state and gapped Dirac surface states

Magnetic topological insulators (TIs) with nontrivial topological electronic structure and broken time-reversal symmetry exhibit various exotic topological quantum phenomena. The realization of such exotic phenomena at high temperature is one of central topics in this area. We reveal that MnBi6Te10 is a magnetic TI with an antiferromagnetic ground state below 10.8 K whose nontrivial topology is manifested by Dirac-like surface states. The ferromagnetic axion insulator state with Z4 = 2 emerges once spins polarized at field as low as 0.1 T, accompanied with saturated anomalous Hall resistivity up to 10 K. Such a ferromagnetic state is preserved even external field down to zero at 2 K. Theoretical calculations indicate that the few-layer ferromagnetic MnBi6Te10 is also topologically nontrivial with a non-zero Chern number. Angle-resolved photoemission spectroscopy experiments further reveal three types of Dirac surface states arising from different terminations on the cleavage surfaces, one of which has insulating behavior with an energy gap of ~ 28 meV at the Dirac point. These outstanding features suggest that MnBi6Te10 is a promising system to realize various topological quantum effects at zero field and high temperature.

cond-mat.mtrl-sci↗