arXiv subjects

Kevin Liang

Publications and source records attributed to Kevin Liang.

12 recordsLinked to original sources

arXiv2026

Scheduling Parallel Optical Circuit Switches for AI Training

The rapid growth of AI training has dramatically increased datacenter traffic demand and energy consumption, which has motivated renewed interest in optical circuit switches (OCSes) as a high-bandwidth, energy-efficient alternative for AI fabrics. Deploying multiple parallel OCSes is a leading alternative. However, efficiently scheduling time-varying traffic matrices across parallel optical switches with non-negligible reconfiguration delays remains an open challenge. We consider the problem of scheduling a single AI traffic demand matrix $D$ over $s$ parallel OCSes while minimizing the makespan under reconfiguration delay $\delta$. Our algorithm Spectra relies on a three-step approach: Decompose $D$ into a minimal set of weighted permutations; Schedule these permutations across parallel switches using load-aware assignment; then Equalize the imbalanced loads on the switches via controlled permutation splitting. Evaluated on realistic AI training workloads (GPT model and Qwen MoE expert routing) as well as standard benchmarks, Spectra vastly outperforms a baseline based on state-of-the-art algorithms, reducing schedule makespan by an average factor of $1.4\times$ on GPT AI workloads, $1.9\times$ on MoE AI workloads, and $2.4\times$ on standard benchmarks. Further, the makespans achieved by Spectra consistently approach newly derived lower bounds.

cs.NI↗

arXiv2025

Super-resolution with Fourier measurements

Resolving sources beyond the diffraction limit is important in imaging, communications, and metrology. Current image-based methods of super-resolution require phase information (either of the source points or an added filter) and perfect alignment with the centroid of the object. Both inhibit the practical application of these methods, as uniform motion and/or relative jitter destroy their assumptions. Here, we show that measuring intensity in the Fourier plane enables super-resolution without any of the issues of image-based methods. We start with the shift-invariance of the Fourier transform and the observation that the two-point position problem ${x_1,x_2}$ in the near field corresponds to the single-point wavenumber problem $k\ =2\pi/(x_2-x_1)$ in the far field. We consider the full range of mutual coherence and show that for fully coherent sources, the Fourier method saturates the quantum limit, i.e. it gives the best possible measurement. Similar results hold for sub-Rayleigh constellations of $N$ sources, which can act collectively as a spatially averaged metasurface and/or individually as elements of a phased-array antenna. The theory paves the way to merge Fourier optics with super-resolution techniques, enabling experimental devices that are both simpler and more robust than previous designs.

physics.optics↗

arXiv2024

OmniPose6D: Towards Short-Term Object Pose Tracking in Dynamic Scenes from Monocular RGB

To address the challenge of short-term object pose tracking in dynamic environments with monocular RGB input, we introduce a large-scale synthetic dataset OmniPose6D, crafted to mirror the diversity of real-world conditions. We additionally present a benchmarking framework for a comprehensive comparison of pose tracking algorithms. We propose a pipeline featuring an uncertainty-aware keypoint refinement network, employing probabilistic modeling to refine pose estimation. Comparative evaluations demonstrate that our approach achieves performance superior to existing baselines on real datasets, underscoring the effectiveness of our synthetic dataset and refinement technique in enhancing tracking precision in dynamic contexts. Our contributions set a new precedent for the development and assessment of object pose tracking methodologies in complex scenes.

cs.CV↗

arXiv2023

EgoTracks: A Long-term Egocentric Visual Object Tracking Dataset

Visual object tracking is a key component to many egocentric vision problems. However, the full spectrum of challenges of egocentric tracking faced by an embodied AI is underrepresented in many existing datasets; these tend to focus on relatively short, third-person videos. Egocentric video has several distinguishing characteristics from those commonly found in past datasets: frequent large camera motions and hand interactions with objects commonly lead to occlusions or objects exiting the frame, and object appearance can change rapidly due to widely different points of view, scale, or object states. Embodied tracking is also naturally long-term, and being able to consistently (re-)associate objects to their appearances and disappearances over as long as a lifetime is critical. Previous datasets under-emphasize this re-detection problem, and their "framed" nature has led to adoption of various spatiotemporal priors that we find do not necessarily generalize to egocentric video. We thus introduce EgoTracks, a new dataset for long-term egocentric visual object tracking. Sourced from the Ego4D dataset, this new dataset presents a significant challenge to recent state-of-the-art single-object tracking models, which we find score poorly on traditional tracking metrics for our new dataset, compared to popular benchmarks. We further show improvements that can be made to a STARK tracker to significantly increase its performance on egocentric data, resulting in a baseline model we call EgoSTARK. We publicly release our annotations and benchmark, hoping our dataset leads to further advancements in tracking.

cs.CV↗

arXiv2022

Off-axis Aberrations Improve the Resolution Limits of Incoherent Imaging

The presence of off-axis tilt and Petzval curvature, two of the lowest-order off-axis Seidel aberrations, is shown to improve the Fisher information of two-point separation estimation in an incoherent imaging system compared to an aberration-free system. Our results show that the practical localization advantages of modal imaging techniques within the field of quantum-inspired superresolution can be achieved with direct imaging measurement schemes alone.

physics.optics↗

arXiv2022

Valley engineering electron-hole liquids in TMDC monolayers

Electron-hole liquids(EHLs), a correlated state of matter and a thermodynamic liquid, have recently been found to exist at room temperature in suspended monolayers of MoS2. Appreciably higher rates of radiative recombination inside the liquid as compared to free excitons hold promise for optoelectronic applications such as broadband lasing. In this paper, we show that leveraging the valley physics in MoS2 may be a route towards achieving tunability of specific characteristics of an EHL, such as emission wavelength, linewidth, and most importantly, the liquid density. The conditions under which EHLs form, in bulk semiconductors as well as TMDC monolayers are quite stringent, requiring high crystal purity and cryogenic temperatures in bulk semiconductors, and suspension in monolayers. Using a simple yet powerful model for describing free excitons and show that a phase transition into the EHL state may be feasible in substrate-supported monolayer samples. More repeatable experimental realizations of EHLs may be essential to answer questions regarding the nature of electron-hole correlations and how they may be used to generate non-trivial states of light.

cond-mat.mtrl-sci↗

arXiv2021

An extended Krylov subspace method for decoding edge-based compressed images by homogeneous diffusion

The heat equation is often used in order to inpaint dropped data in inpainting-based lossy compression schemes. We propose an alternative way to numerically solve the heat equation by an extended Krylov subspace method. The method is very efficient with respect to the direct computation of the solution of the heat equation at large times. And this is exactly what is needed for decoding edge-compressed pictures by homogeneous diffusion.

cs.IT↗

arXiv2021

Coherence effects on estimating general sub-Rayleigh object distribution moments

Quantum Fisher information and signal-to-noise ratio bounds are derived for the estimation of moments of general partially coherent objects. Under an asymptotic analysis in the sub-Rayleigh regime, these bounds are shown to be less restrictive than those of the incoherent case. This leaves open the possibility of more accurately estimating object distribution moments when the object is partially coherent.

physics.optics↗

arXiv2021

Meta-Learned Attribute Self-Gating for Continual Generalized Zero-Shot Learning

Zero-shot learning (ZSL) has been shown to be a promising approach to generalizing a model to categories unseen during training by leveraging class attributes, but challenges still remain. Recently, methods using generative models to combat bias towards classes seen during training have pushed the state of the art of ZSL, but these generative models can be slow or computationally expensive to train. Additionally, while many previous ZSL methods assume a one-time adaptation to unseen classes, in reality, the world is always changing, necessitating a constant adjustment for deployed models. Models unprepared to handle a sequential stream of data are likely to experience catastrophic forgetting. We propose a meta-continual zero-shot learning (MCZSL) approach to address both these issues. In particular, by pairing self-gating of attributes and scaled class normalization with meta-learning based training, we are able to outperform state-of-the-art results while being able to train our models substantially faster ($>100\times$) than expensive generative-based approaches. We demonstrate this by performing experiments on five standard ZSL datasets (CUB, aPY, AWA1, AWA2 and SUN) in both generalized zero-shot learning and generalized continual zero-shot learning settings.

cs.CV↗

arXiv2021

Experimental demonstration of superresolution of partially coherent light sources using parity sorting

Analyses based on quantum metrology have shown that the ability to localize the positions of two incoherent point sources can be significantly enhanced through the use of mode sorting. Here we theoretically and experimentally investigate the effect of partial coherence on the sub-diffraction limit localization of two sources based on parity sorting. With the prior information of a negative and real-valued degree of coherence, higher Fisher information is obtained than that for the incoherent case. Our results pave the way to clarifying the role of coherence in quantum limited metrology.

physics.optics↗

arXiv2020

Coherence effects on estimating two-point separation

The quantum Fisher information (FI), when applied to the estimation of the separation of two point sources, has been shown to be non-zero in cases where the coherence between the sources are known. Although it has been claimed that ignorance of the coherence causes the quantum FI to vanish (a resurgence of Rayleigh's curse), a more complete analysis including both the magnitude and phase of the coherence parameter is given here. Partial ignorance of the coherence is shown to potentially break Rayleigh's curse, whereas complete ignorance guarantees its resurgence.

physics.optics↗

arXiv2019

Theoretical analysis of quantum random walks with stress-engineered optics

Quantum random walks (QRWs) are random processes in which the resulting probability density of the "walker" state, whose movement is governed by a "coin" state, is described in a non-classical manner. Previously, Q-plates have been used to demonstrate QRWs with polarization and orbital angular momentum playing the role of coin and walker states, respectively. In this theoretical analysis, we show how stress-engineered optics can be used to develop new platforms for complex QRWs through relative simple optical elements. Our work opens up new paths to speed up classical-to-quantum transitions in robust photonic networks.

physics.optics↗