Search

Momentum-based optimizers are widely adopted for training neural networks. However, the optimal selection of momentum coefficients remains elusive. This uncertainty impedes a clear understanding of the role of momentum in stochastic gradient methods. In this paper, we present a frequency domain analysis framework that interprets the momentum method as a time-variant filter for gradients, where adjustments to momentum coefficients modify the filter characteristics. Our experiments support this perspective and provide a deeper understanding of the mechanism involved. Moreover, our analysis reveals the following significant findings: high-frequency gradient components are undesired in the late stages of training; preserving the original gradient in the early stages, and gradually amplifying low-frequency gradient components during training both enhance generalization performance. Based on these insights, we propose Frequency Stochastic Gradient Descent with Momentum (FSGDM), a heuristic optimizer that dynamically adjusts the momentum filtering characteristic with an empirically effective dynamic magnitude response. Experimental results demonstrate the superiority of FSGDM over conventional momentum optimizers.

This paper gives the commensurability classification of Deligne-Mostow ball quotients and shows that the 104 Deligne-Mostow lattices form 38 commensurability classes. Firstly, we find commensurability relations among Deligne-Mostow monodromy groups, which are not necessarily discrete. This recovers and generalizes previous work by Sauter and Deligne-Mostow in dimension two. In this part, we consider certain projective surfaces with two fibrations over the projective line, which induce two sets of Deligne-Mostow data. The correspondences of moduli spaces provide the geometric realization of commensurability relations. Secondly, we obtain commensurability invariants from conformal classes of Hermitian forms and toroidal boundary divisors. This completes the commensurability classification of Deligne-Mostow lattices and also reproves Kappes-Moller and McMullen's results on non-arithmetic Deligne-Mostow lattices., Comment: 46 pages, 14 figures, comments welcome!

Realistic audio synthesis that captures accurate acoustic phenomena is essential for creating immersive experiences in virtual and augmented reality. Synthesizing the sound received at any position relies on the estimation of impulse response (IR), which characterizes how sound propagates in one scene along different paths before arriving at the listener's position. In this paper, we present Acoustic Volume Rendering (AVR), a novel approach that adapts volume rendering techniques to model acoustic impulse responses. While volume rendering has been successful in modeling radiance fields for images and neural scene representations, IRs present unique challenges as time-series signals. To address these challenges, we introduce frequency-domain volume rendering and use spherical integration to fit the IR measurements. Our method constructs an impulse response field that inherently encodes wave propagation principles and achieves state-of-the-art performance in synthesizing impulse responses for novel poses. Experiments show that AVR surpasses current leading methods by a substantial margin. Additionally, we develop an acoustic simulation platform, AcoustiX, which provides more accurate and realistic IR simulations than existing simulators. Code for AVR and AcoustiX are available at https://zitonglan.github.io/avr., Comment: NeurIPS 2024 Spotlight

In 2009, Dolgachev-Keum show that finite groups of tame symplectic automorphisms of K3 surfaces in positive characteristics are subgroups of the Mathieu group of degree 23 under mild conditions. In this paper, we utilize lattice-theoretic methods to investigate symplectic actions of finite groups G on K3 surfaces in odd characteristics. In the tame cases (i.e., the order of G is coprime with p) and all the superspecial cases, one can associate with the action a Leech pair which can be detected via H\"ohn-Mason's list. Notice that the superspecial cases have been resolved by Ohashi-Sch\"utt. If the K3 surface is supersingular with Artin invariant at least two, we develop a new machinery called p-root pairs to detect possible symplectic finite group actions (without the assumption of tameness). The concept of p-root pair is closely related to root systems and Weyl groups. In particular, we recover many results by Dolgachev-Keum with a different method and give an upper bound for the exponent of p in |G|., Comment: Any comments are welcome

This article theoretically proposes a multi-layer Fabry-Perot cavity structure based on nonlinear MoS2, whose cavity is composed of asymmetric photonic crystals. In this structure, we observed a low threshold optical bistability phenomenon on the order of a in the visible light band, which is caused by the large third-order nonlinear conductivity of the bilayer MoS2 and the Fabry-Perot cavity resonance. Research has found that when light is incident from two different directions in an asymmetric Fabry-Perot cavity, the optical bistability exhibits not exactly the same behavior. In addition, we further investigated and found that the optical bistability behavior in this simple multi-layer structure is closely related to parameters such as incident wavelength, Fabry-Perot cavity length, and refractive index of the photonic crystal dielectric. This work provides a new approach for the implementation of low threshold optical bistable devices in the visible light band, which is expected to be applied in nonlinear optical fields such as all optical switches and all optical logic devices., Comment: 22 pages, 6 figures

In this paper, we present a theoretical analysis of the optical bistability in a metallic silver-Bragg reflector structure by embedding bilayer MoS2 at the visible band. The nonlinear OB is achieved due to the nonlinear conductivity of the bilayer MoS2 and the excitation of the optical Tamm state at the interface between the silver and the Bragg reflector. It is found that the hysteresis behaviour and the threshold width of the OB can be effectively tuned by varying the incident light wavelength. In addition, the optical bistable behaviour of the structure can be adjusted by varying the position of the MoS2 inset in the defect layer, incident angle and the structural parameters of the spacer layer. Although the current threshold cannot be commercialized, we believe that this solution will provide a meaningful path reference for low threshold bistability in the visible light band., Comment: I am deeply sorry to submit a request for the withdrawal of our paper to you esteemed journal.We believe that there are some areas in the manuscript that require further improvement,and out of our sense of responsiblity towards your journal,we have decided to withdraw our paper after caerful deliberation and discussion

Although supervised machine learning is popular for information extraction from clinical notes, creating large annotated datasets requires extensive domain expertise and is time-consuming. Meanwhile, large language models (LLMs) have demonstrated promising transfer learning capability. In this study, we explored whether recent LLMs can reduce the need for large-scale data annotations. We curated a manually-labeled dataset of 769 breast cancer pathology reports, labeled with 13 categories, to compare zero-shot classification capability of the GPT-4 model and the GPT-3.5 model with supervised classification performance of three model architectures: random forests classifier, long short-term memory networks with attention (LSTM-Att), and the UCSF-BERT model. Across all 13 tasks, the GPT-4 model performed either significantly better than or as well as the best supervised model, the LSTM-Att model (average macro F1 score of 0.83 vs. 0.75). On tasks with high imbalance between labels, the differences were more prominent. Frequent sources of GPT-4 errors included inferences from multiple samples and complex task design. On complex tasks where large annotated datasets cannot be easily collected, LLMs can reduce the burden of large-scale data labeling. However, if the use of LLMs is prohibitive, the use of simpler supervised models with large annotated datasets can provide comparable results. LLMs demonstrated the potential to speed up the execution of clinical NLP studies by reducing the need for curating large annotated datasets. This may result in an increase in the utilization of NLP-based variables and outcomes in observational clinical studies.

In this paper we mainly study Calabi-Yau varieties that arise as triple covers of products of projective lines branched along simple normal crossing divisors. For some of those families of Calabi-Yau varieties, the period maps factor through arithmetic quotients of complex hyperbolic balls. We give a classification of such examples. One of the families was previously studied by Voisin, Borcea and Rohde. For these ball-type cases, we will show arithmeticity of the monodromy groups. These ball quotients are all commensurable to ball quotients in Deligne-Mostow theory. As a byproduct, we prove some commensurability relations among arithmetic groups in Deligne-Mostow theory., Comment: 34 pages, comments welcome

Federated learning (FL) enables multiple clients to collaboratively train deep learning models while considering sensitive local datasets' privacy. However, adversaries can manipulate datasets and upload models by injecting triggers for federated backdoor attacks (FBA). Existing defense strategies against FBA consider specific and limited attacker models, and a sufficient amount of noise to be injected only mitigates rather than eliminates FBA. To address these deficiencies, we introduce a Flexible Federated Backdoor Defense Framework (Fedward) to ensure the elimination of adversarial backdoors. We decompose FBA into various attacks, and design amplified magnitude sparsification (AmGrad) and adaptive OPTICS clustering (AutoOPTICS) to address each attack. Meanwhile, Fedward uses the adaptive clipping method by regarding the number of samples in the benign group as constraints on the boundary. This ensures that Fedward can maintain the performance for the Non-IID scenario. We conduct experimental evaluations over three benchmark datasets and thoroughly compare them to state-of-the-art studies. The results demonstrate the promising defense performance from Fedward, moderately improved by 33% $\sim$ 75 in clustering defense methods, and 96.98%, 90.74%, and 89.8% for Non-IID to the utmost extent for the average FBA success rate over MNIST, FMNIST, and CIFAR10, respectively., Comment: Accepted by IEEE ICME 2023

The secant variety of the Veronese surface is a singular cubic fourfold. The degeneration of Hodge structures of one-parameter degenerations to this secant cubic fourfold is a key ingredient for B.~Hassett and R.~Laza in studying the moduli space of cubic fourfolds via the period mapping. We generalize some of their results to the cubic hypersurface that is the secant variety of a Severi variety. Specifically, we study the limit mixed Hodge structures associated to one-parameter degenerations to the secant cubic hypersurface. Considering S.~Usui's partial compactification of a period domain for Hodge structures of general weights, we apply the limit mixed Hodge structure to characterize a local extension of the period map for the corresponding cubic hypersurfaces., Comment: 34 pages. Some typos are fixed. Comments welcome

We study moduli spaces of certain sextic curves with a singularity of multiplicity 3 from both perspectives of Deligne-Mostow theory and periods of K3 surfaces. In both ways we can describe the moduli spaces via arithmetic quotients of complex hyperbolic balls. We show in Theorem 7.4 that the two ball-quotient constructions can be unified in a geometric way., Comment: 26 pages, comments welcome!

Metasurface-mediated bound states in the continuum (BIC) provides a versatile platform for light manipulation at subwavelength dimension with diverging radiative quality factor and extreme optical localization. In this work, we employ magnetic dipole quasi-BIC resonance in asymmetric silicon nanobar metasurfaces to realize giant Goos-H\"anchen (GH) shift enhancement by more than three orders of wavelength. In sharp contrast to GH shift based on the Brewster dip or transmission-type resonance, the maximum GH shift here is located at the reflection peak with unity reflectance, which can be conveniently detected in the experiment. By adjusting the asymmetric parameter of metasurfaces, the $Q$-factor and GH shift can be modulated accordingly. More interestingly, it is found that GH shift exhibits an inverse quadratic dependence on the asymmetric parameter. Furthermore, we design an ultrasensitive environmental refractive index sensor based on the quasi-BIC enhanced GH shift, with a maximum sensitivity of 1.5$\times$10$^{7}$ $\mu$m/RIU. Our work not only reveals the essential role of BIC in engineering the basic optical phenomena, but also suggests the way for pushing the performance limits of optical communication devices, information storage, wavelength division de/multiplexers, and ultrasensitive sensors.

In this paper, we develop a new method to classify abelian automorphism groups of hypersurfaces. We use this method to classify (Theorem 4.2) abelian groups that admit a liftable action on a smooth cubic fourfold. A parallel result (Theorem 5.1) is obtained for quartic surfaces., Comment: 16 pages, comments welcome!