arXiv subjects

Jason Tang

Publications and source records attributed to Jason Tang.

8 recordsLinked to original sources

arXiv2026

Waves dictate the yo-yoing decay of a viscoelastic mixing layer

We find that waves develop in a time-decaying mixing layer of viscoelastic fluid, leading the mean-flow to yo-yo. This is in sharp contrast with Newtonian fluids, where laminar mixing layers evolve monotonically. We combine direct numerical simulations with a theoretical analysis of the energy budget for the flow to uncover the underlying physical mechanism. The yo-yoing of the mean-flow is shown to be driven by the elastic polymers injecting energy into the fluid and, in turn, being rotated by the large-scale mean shear. We then provide the mathematical model of the problem and solve it analytically, finding wave solutions with non-linear dispersion predicting the period of the yo-yoing and the parameter range where it occurs. As decaying mixing layers are one of the simplest and canonical examples of unsteady flows, the phenomenon identified here explains the anomalies recently observed in experiments of unsteady viscoelastic flows in complex geometries.

physics.flu-dyn↗

arXiv2026

How Many Visual Levers Drive Urban Perception? Interventional Counterfactuals via Multiple Localised Edits

Street-view perception models predict subjective attributes such as safety at scale, but remain correlational: they do not identify which localized visual changes would plausibly shift human judgement for a specific scene. We propose a lever-based interventional counterfactual framework that recasts scene-level explainability as a bounded search over structured counterfactual edits. Each lever specifies a semantic concept, spatial support, intervention direction, and constrained edit template. Candidate edits are generated through prompt-conditioned image editing and retained only if they satisfy validity checks for same-place preservation, locality, realism, and plausibility. In a pilot across 50 scenes from five cities, the framework reveals preliminary proxy-based directional patterns and a practical failure taxonomy under prompt-only editing, with Mobility Infrastructure and Physical Maintenance showing the largest auxiliary safety shifts. Human pairwise judgements remain the ground-truth endpoint for future validation.

cs.CY↗

arXiv2024

Analyzing Customer-Facing Vendor Experiences with Time Series Forecasting and Monte Carlo Techniques

eBay partners with external vendors, which allows customers to freely select a vendor to complete their eBay experiences. However, vendor outages can hinder customer experiences. Consequently, eBay can disable a problematic vendor to prevent customer loss. Disabling the vendor too late risks losing customers willing to switch to other vendors, while disabling it too early risks losing those unwilling to switch. In this paper, we propose a data-driven solution to answer whether eBay should disable a problematic vendor and when to disable it. Our solution involves forecasting customer behavior. First, we use a multiplicative seasonality model to represent behavior if all vendors are fully functioning. Next, we use a Monte Carlo simulation to represent behavior if the problematic vendor remains enabled. Finally, we use a linear model to represent behavior if the vendor is disabled. By comparing these forecasts, we determine the optimal time for eBay to disable the problematic vendor.

stat.ML↗

arXiv2024

Shopping Queries Image Dataset (SQID): An Image-Enriched ESCI Dataset for Exploring Multimodal Learning in Product Search

Recent advances in the fields of Information Retrieval and Machine Learning have focused on improving the performance of search engines to enhance the user experience, especially in the world of online shopping. The focus has thus been on leveraging cutting-edge learning techniques and relying on large enriched datasets. This paper introduces the Shopping Queries Image Dataset (SQID), an extension of the Amazon Shopping Queries Dataset enriched with image information associated with 190,000 products. By integrating visual information, SQID facilitates research around multimodal learning techniques that can take into account both textual and visual information for improving product search and ranking. We also provide experimental results leveraging SQID and pretrained models, showing the value of using multimodal data for search and ranking. SQID is available at: https://github.com/Crossing-Minds/shopping-queries-image-dataset.

cs.IR↗

arXiv2024

Captions Are Worth a Thousand Words: Enhancing Product Retrieval with Pretrained Image-to-Text Models

This paper explores the usage of multimodal image-to-text models to enhance text-based item retrieval. We propose utilizing pre-trained image captioning and tagging models, such as instructBLIP and CLIP, to generate text-based product descriptions which are combined with existing text descriptions. Our work is particularly impactful for smaller eCommerce businesses who are unable to maintain the high-quality text descriptions necessary to effectively perform item retrieval for search and recommendation use cases. We evaluate the searchability of ground-truth text, image-generated text, and combinations of both texts on several subsets of Amazon's publicly available ESCI dataset. The results demonstrate the dual capability of our proposed models to enhance the retrieval of existing text and generate highly-searchable standalone descriptions.

cs.IR↗

arXiv2022

Performance of the X-Calibur Hard X-Ray Polarimetry Mission during its 2018/19 Long-Duration Balloon Flight

X-Calibur is a balloon-borne telescope that measures the polarization of high-energy X-rays in the 15--50keV energy range. The instrument makes use of the fact that X-rays scatter preferentially perpendicular to the polarization direction. A beryllium scattering element surrounded by pixellated CZT detectors is located at the focal point of the InFOC{\mu}S hard X-ray mirror. The instrument was launched for a long-duration balloon (LDB) flight from McMurdo (Antarctica) on December 29, 2018, and obtained the first constraints of the hard X-ray polarization of an accretion-powered pulsar. Here, we describe the characterization and calibration of the instrument on the ground and its performance during the flight, as well as simulations of particle backgrounds and a comparison to measured rates. The pointing system and polarimeter achieved the excellent projected performance. The energy detection threshold for the anticoincidence system was found to be higher than expected and it exhibited unanticipated dead time. Both issues will be remedied for future flights. Overall, the mission performance was nominal, and results will inform the design of the follow-up mission XL-Calibur, which is scheduled to be launched in summer 2022.

astro-ph.IM↗

arXiv2021

Cadmium Zinc Telluride Detectors for a Next-Generation Hard X-ray Telescope

We are currently developing Cadmium Zinc Telluride (CZT) detectors for a next-generation space-borne hard X-ray telescope which can follow up on the highly successful NuSTAR (Nuclear Spectroscopic Telescope Array) mission. Since the launch of NuSTAR in 2012, there have been major advances in the area of X-ray mirrors, and state-of-the-art X-ray mirrors can improve on NuSTAR's angular resolution of ~1 arcmin Half Power Diameter (HPD) to 15" or even 5" HPD. Consequently, the size of the detector pixels must be reduced to match this resolution. This paper presents detailed simulations of relatively thin (1 mm thick) CZT detectors with hexagonal pixels at a next-neighbor distance of 150 $\mu$m. The simulations account for the non-negligible spatial extent of the deposition of the energy of the incident photon, and include detailed modeling of the spreading of the free charge carriers as they move toward the detector electrodes. We discuss methods to reconstruct the energies of the incident photons, and the locations where the photons hit the detector. We show that the charge recorded in the brightest pixel and six adjacent pixels suffices to obtain excellent energy and spatial resolutions. The simulation results are being used to guide the design of a hybrid application-specific integrated circuit (ASIC)-CZT detector package.

astro-ph.IM↗

arXiv2020

Few distance sets in $\ell_p$ spaces and $\ell_p$ product spaces

Kusner asked if $n+1$ points is the maximum number of points in $\mathbb{R}^n$ such that the $\ell_p$ distance $(1<p<\infty)$ between any two points is $1$. We present an improvement to the best known upper bound when $p$ is large in terms of $n$, as well as a generalization of the bound to $s$-distance sets. We also study equilateral sets in the $\ell_p$ sums of Euclidean spaces, deriving upper bounds on the size of an equilateral set for when $p=\infty$, $p$ is even, and for any $1\le p<\infty$.

math.MG↗