arXiv Science⌕ Search

arXiv subjects

Hongwei Ding

Publications and source records attributed to Hongwei Ding.

6 recordsLinked to original sources

InertiaMPDD: Multimodal Personalized Depression Detection via Emotional Inertia-Inspired Dynamics Modeling

Multimodal depression screening estimates risk from speech, visual behavior and language. Current methods mainly predict labels, with limited evidence of departures from a non-depressed reference. Recording-level supervision also leaves fine-grained emotional dynamics unannotated. Motivated by emotional inertia, the persistence of affect, InertiaMPDD learns healthy next-state patterns using a world-model-inspired latent predictor. Using pretrained multimodal representations, it predicts from observed history and context. Reference deviation, residual decay rate and a Jacobian spectral descriptor form reference-derived emotional dynamics descriptors, evaluated as candidate proxy biomarkers for standalone screening and personalized decision correction. Across five cohorts from four datasets, fusion reaches 98.41% F1 on D-Vlog and 84.98% on LMVD. Standalone descriptors outperform fixed additive fusion on MPDD but discriminate weakly on AVEC2014. These findings distinguish reference informativeness from fusion benefit and reveal context-dependent utility.

cs.AI↗

Two Routes to the Middle: Placement Search and Brain Readouts Converge on Where Continual Learners Should Specialize

Continual learners that keep a task-specific adapter in every block of a pre-trained vision transformer accumulate storage linearly with the number of tasks; keeping task-specific adapters in only a few blocks curbs this growth but raises the question of where to place them. We investigate this question from two perspectives. Algorithmically, training all contiguous four-block placements yields an inverted U: final accuracy peaks at intermediate depth and varies by up to 3.5 percentage points (pp), while inexpensive criteria based on weight spectra or activation statistics favor the deepest blocks. From neuroscience, the hierarchical organization and intermediate-stage plasticity of the visual cortex motivate us to ask whether a measurement taken outside the learner can guide layer specialization without placement search. LS-B observes the first tasks through a frozen fMRI encoding model of twelve human visual areas and commits task-specific capacity once to the blocks whose readouts vary most across tasks relative to their stable structure. Across three ViT-B/16 backbones, LS-B yields stable, backbone-specific allocations. On the two backbones with placement search, AugReg and iBOT, the selected blocks overlap the intermediate-depth region identified by search. Under matched storage and observation budgets, the selected blocks outperform the shallowest and deepest four-block configurations. On Split ImageNet-R, LS-B uses 60% of full-BiLoRA adapter storage while remaining within 1.5 pp of its final accuracy. The allocation requires no labels or backpropagation, adds under 0.6% runtime, and exhibits backbone-specific cortical signatures.

cs.LG↗

RBE-Flow: Recurrent Bayesian Estimation on Feature Manifolds for Cross-Modal Registration

Cross-modal image registration is essential for multi-sensor perception but remains fundamentally challenging due to severe non-linear radiometric discrepancies and geometric distortions. Existing deterministic matching methods lack uncertainty awareness, struggling to navigate the resulting highly non-convex optimization landscape and frequently accumulating errors in ambiguous regions. In this paper, we propose RBE-Flow, a novel framework that reformulates dense cross-modal flow estimation as a closed-loop recurrent Bayesian estimation problem on learned feature manifolds. Diverging from standard feed-forward regression, RBE-Flow establishes a robust self-correcting mechanism by deeply coupling feature-metric non-linear optimization with probabilistic state updates. Specifically, a Recurrent Manifold Optimization (RMO) block iteratively generates flow observations and their associated uncertainties, which are then optimally assimilated into the prior state via an Uncertainty-Adaptive Probabilistic Update (UAPU) using deterministic sigma-point projection. Crucially, the resulting calibrated posterior covariance is fed back to adaptively regularize the damping of subsequent optimization steps, allowing the system to modulate its convergence based on predictive confidence. To ensure stable probabilistic training, we introduce a hybrid supervision scheme featuring a geometry-aware rectified NLL loss that structurally prevents variance collapse. Extensive experiments on challenging OSdataset, WHU-OPT-SAR, and RoadScene benchmarks demonstrate that RBE-Flow consistently achieves state-of-the-art performance, outperforming existing methods by a significant margin, particularly under strict sub-pixel criteria. Project page: https://github.com/NEU-Liuxuecong/RBE-Flow

cs.CV↗

Polarization-independent resonant lattice Kerker effect in phase-change metasurface

Resonant lattice Kerker effect in periodic resonators is one of the most interesting generalizations of the Kerker effect that relates to various fascinating functionalities such as scattering management and Huygens metasurfaces. However, so far this effect has been shown to be sensitive to the incident polarization, restricting its applications. Here, we report, for the first time, polarization-independent resonant lattice Kerker effect in metasurfaces composed of periodic Ge$_2$Se$_2$Te$_5$ (GST) disks. For such a metasurface of square lattice, the spectrally overlap of the electric dipole and magnetic dipole surface lattice resonances can be realized by choosing an appropriate GST crystalline fraction regardless of the incident polarization. The operation wavelength and the required GST crystalline fraction can be conveniently tuned over large ranges since these parameters scale linearly with the disk size and the lattice period, greatly facilitating the design. Making use of the obtained resonant lattice Kerker effect, we realize a reconfigurable and polarization-independent lattice Huygens' metasurface with a dynamic phase modulation of close to $2π$ and high transmittance. This work will advance the engineering of the resonant lattice Kerker effect and promote its applications in phase modulation and wavefront control.

physics.optics↗

Active Tuning of Resonant Lattice Kerker Effect

The Kerker effect has been generalized in nanophotonics and meta-optics, and has recently been of great interest by relating to various fascinating functionalities such as scattering management and perfect transmission, reflection or absorption. One of the most interesting generalizations is the resonant lattice Kerker effect in periodic nanostructures. However, its active tuning has not been explored yet. Here, we report, for the first time, the active control of the resonant lattice Kerker effect in periodic Ge2Se2Te5 nanodisks. By changing the crystalline fraction, we show that the electric dipole lattice resonance (ED-LR), the magnetic dipole resonance (MDR), and thus the resonant lattice Kerker effect are all red-shifted. We therefore realize the transition from the ED-LR to the resonant lattice Kerker effect, which enables multilevel tuning of reflection, transmission and absorption with modulation depths above 86%. Taking advantage of the MDR redshifts, we also observe broadband and multilevel tuning of transmission with modulation depth of 87% over a broadband range of 588 nm. Our work establishes a new path for designing high-performance active nanophotonic devices.

physics.optics↗

Narrow quadrupolar surface lattice resonances and band reversal in vertical metal-insulator-metal gratings

We report narrow quadrupolar surface lattice resonances (SLRs) under normal incidence, and the observation, for the first time, of the band reversal effect of SLRs supported by a vertical metal-insulator-metal nanograting, which is embedded in a homogeneous dielectric environment. Simulation results show that under normal incidence, quadrupolar SLR with linewidth of 1~nm and high quality factor of 979 can be excited in the near-infrared regime, and that under oblique incidence, out-of-plane dipolar SLRs of relatively large quality factors (>=150) can be launched. By varying the incidence angle, the SLR wavelength can be continuously tuned over an extremely broadband range of 750 nm, covering most of the near-infrared regime, and the quality factor decreases exponentially. Remarkably, the resonance lineshape can also be dynamically tuned from an asymmetric Fano-shaped dip to a peak, a dip/peak pair, and a perfect symmetric Lorentzian peak, suggesting the appearance of the band reversal effect. We expect the high-Q SLRs with broadband tunability and tunable lineshapes will find potential applications in enhanced nanoscale light-matter interactions in nanolasers, nonlinear optics and sensing.

physics.optics↗