A5100345589- Quantum optics and atomic interactions
- Privacy-Preserving Technologies in Data
- Quantum Information and Cryptography
- Magnetic properties of thin films
- Cold Atom Physics and Bose-Einstein Condensates
- Stochastic Gradient Optimization Techniques
- Atomic and Subatomic Physics Research
- Advanced Fluorescence Microscopy Techniques
- Mobile Crowdsensing and Crowdsourcing
- Machine Learning and ELM
- Magneto-Optical Properties and Applications
- Laser-Matter Interactions and Applications
- Data Management and Algorithms
- Nonlinear Optical Materials Studies
- Advanced Database Systems and Queries
- Advanced Fiber Laser Technologies
- Quantum Mechanics and Applications
- Advanced Computational Techniques and Applications
- Optical Coherence Tomography Applications
- Optical Network Technologies
- Advanced Optical Imaging Technologies
- Plant Micronutrient Interactions and Effects
- Topic Modeling
- Advanced Wireless Communication Technologies
- Advanced Fiber Optic Sensors
University of Tennessee at Chattanooga
2022-2025
Kyoto University Institute for Chemical Research
2017-2025
Kyoto University
2017-2025
Kyoto Bunkyo University
2017-2025
First Affiliated Hospital of Bengbu Medical College
2024
Hainan University
2024
Chinese University of Hong Kong
2024
Wuhan Institute of Technology
2023
Shanxi University
2023
State Key Laboratory of Quantum Optics and Quantum Optics Devices
2023
Federated Learning is a distributed learning paradigm with two key challenges that differentiate it from traditional optimization: (1) significant variability in terms of the systems characteristics on each device network (systems heterogeneity), and (2) non-identically data across (statistical heterogeneity). In this work, we introduce framework, FedProx, to tackle heterogeneity federated networks. FedProx can be viewed as generalization re-parametrization FedAvg, current state-of-the-art...
Federated learning involves training statistical models over remote devices or siloed data centers, such as mobile phones hospitals, while keeping localized. Training in heterogeneous and potentially massive networks introduces novel challenges that require a fundamental departure from standard approaches for large-scale machine learning, distributed optimization, privacy-preserving analysis. In this article, we discuss the unique characteristics of federated provide broad overview current...
Modern federated networks, such as those comprised of wearable devices, mobile phones, or autonomous vehicles, generate massive amounts data each day. This wealth can help to learn models that improve the user experience on device. However, scale and heterogeneity presents new challenges in research areas learning, meta-learning, multi-task learning. As machine learning community begins tackle these challenges, we are at a critical time ensure developments made grounded with realistic...
Federated learning aims to jointly learn statistical models over massively distributed remote devices. In this work, we propose FedDANE, an optimization method that adapt from DANE [8], [9], a for classical optimization, handle the practical constraints of federated learning. We provide convergence guarantees when both convex and non-convex functions. Despite encouraging theoretical results, find has underwhelming performance empirically. particular, through empirical simulations on...
The nonreciprocity of propagating spin waves, i.e., the difference in amplitude and/or frequency depending on propagation direction, is essential for realization wave-based logic circuits. However, nonreciprocal shifts demonstrated so far are not large enough applications because they originate from interfacial effects. In addition, switching wave electrical way remains a challenging issue. Here, we show switchable giant shift waves interlayer exchange-coupled synthetic antiferromagnets....
Brillouin microscopy is an emerging label-free imaging technique used to assess local viscoelastic properties. Quantum-enhanced stimulated scattering demonstrated using low power continuous-wave lasers at 795 nm. A signal-to-noise ratio enhancement of 3.4 dB reported by two-mode intensity-difference squeezed light generated with the four-wave mixing process in atomic rubidium vapor. The optical and excitation wavelengths water transparency window have potential provide a powerful bio-imaging...
In this work, we explore the unique challenges -- and opportunities of unsupervised federated learning (FL). We develop analyze a one-shot clustering scheme, $k$-FED, based on widely-used Lloyd's method for $k$-means clustering. contrast to many supervised problems, show that issue statistical heterogeneity in networks can fact benefit our analysis. analyse $k$-FED under center separation assumption compare it best known requirements its centralized counterpart. Our analysis shows...
Distributed optimization has become the default training paradigm in modern machine learning due to growing scale of models and datasets. To mitigate communication overhead, local updates are often applied before global aggregation, resulting a nested approach with inner outer steps. However, heavy-tailed stochastic gradient noise remains significant challenge, particularly attention-based models, hindering effective training. In this work, we propose TailOPT, an efficient framework designed...
Background Kunitz-type venom peptides have been isolated from a wide variety of venomous animals. They usually protease inhibitory activity or potassium channel blocking activity, which by virtue the effects on predator animals are essential for survival However, no scorpion functionally characterized. Principal Findings A new peptide gene precursor, SdPI, was cloned and characterized gland cDNA library Lychas mucronatus. It codes signal 21 residues mature 59 residues. The SdPI possesses...
All-optical helicity dependent switching (AO-HDS), deterministic control of magnetization by circularly polarized laser pulses, allows to efficiently manipulate spins without the need a magnetic field. However, AO-HDS in ferromagnetic metals so far requires many pulses for fully their states. Using combination short, 90-fs linearly pulse and subsequent longer, 3-ps pulse, we demonstrate that number full reversal can be reduced 4 pairs single stack Pt/Co/Pt. The obtained results suggest...
We present ease.ml, a declarative machine learning service platform. With user defines the high-level schema of an ML application and submits task via Web interface. The system then deals with rest, such as model selection data movement. ultimate question we hope to understand is that, "service provider" that manages shared cluster machines running workloads, what resource sharing strategy maximizes global satisfaction all our users?
Absorption measurement is an exceptionally versatile tool for many applications in science and engineering. For absorption measurements using laser beams of light, the sensitivity theoretically limited by shot noise due to fundamental Poisson distribution photon number radiation. In practice, shot-noise limit can only be achieved when all other sources are eliminated. Here, we use seeded squeezed light demonstrate that direct-absorption performed with a beyond limit. We present practically...
An X-ray tomographic technique was developed to investigate the internal magnetic domain structure in a micrometer-sized ferromagnetic sample. The is based on scanning hard nanoprobe using circular dichroism (XMCD). From transmission XMCD images at Gd L3 edge as function of sample rotation angle, three-dimensional (3D) distribution single component vector GdFeCo microdisc reconstructed with spatial resolution 360 nm, modified algebraic reconstruction algorithm. method applicable practical...
We use the quantum correlations of twin beams light to investigate fundamental addition noise when one propagates through a fast-light medium based on phase-insensitive gain. The experiment is two successive four-wave mixing processes in rubidium vapor, which allow for generation bright two-mode-squeezed followed by controlled advancement while maintaining shared between beams. demonstrated effect allows study irreversible decoherence exhibiting anomalous dispersion, and first time shows...
Strong quantum correlated sources are essential but delicate resources for information science and engineering protocols. Decoherence loss the two main disruptive processes that lead to of nonclassical behavior in correlations. In systems, scattering can contribute both decoherence loss. this work, we present an experimental scheme capable significantly mitigating adverse impact systems. Our system is composed a two-mode squeezed light generated with four-wave-mixing process hot rubidium...
We demonstrate the ability of a phase-sensitive amplifier (PSA) to pre-amplify selected quadrature one mode two-mode squeezed state in order improve measurement quantum correlations that exist before degradation due optical and detection losses. use four-wave mixing (4WM) $^{85}$Rb vapor generate bright beams state. One these two modes then passes through second 4WM interaction PSA configuration noiselessly desired loss is intentionally introduced. an enhancement measured degree intensity...
We investigate temperature, temporal and magnetic-field dependence of unidirectional magnetoresistance (UMR) in metallic bilayers. The UMR is found to decrease rapidly with reducing temperature converges a finite value at low-temperature limit. shows that the emerges nanosecond time scale, which depends on current amplitude. almost constant up 5 T. These experimental results imply high-energy magnons, are not considered existing theory, can be involved observed UMR. Our therefore suggest...
Two-photon absorption (TPA) fluorescence of biomarkers has been decisive in advancing the fields biosensing and deep-tissue vivo imaging live specimens. However, due to extremely small TPA cross section quadratic dependence on input photon flux, high peak-intensity pulsed lasers are imperative, which can result significant photo- thermal-damage. Previous works entangled (ETPA) with spontaneous parametric down-conversion (SPDC) light sources found a linear photon-pair but limited by low...