arXiv subjects

Yuzhou Peng

Publications and source records attributed to Yuzhou Peng.

5 recordsLinked to original sources

arXiv2024

A second order Langevin sampler preserving positive volume for isothermal isobaric ensemble

We propose in this work a second-order Langevin sampler for the isothermal-isobaric ensemble (the NPT ensemble), preserving a positive volume for the simulation box. We first derive the suitable equations of motion for particles to be coupled with the overdamped Langevin equation of volume by sending the artificial mass of the periodic box to zero in the work of Liang et. al. [J. Chem. Phys. 157(14)]. We prove the well-posedness of the new system of equations and show that its invariant measure is the desired ensemble. The new continuous time equations not only justify the previous cell-rescaling methods, but also allow us to choose a suitable friction coefficient so that one has additive noise after a change of variable by taking logarithm of the volume. This observation allows us to propose a second order weak scheme that guarantees the positivity of the volume. Various numerical experiments have been performed to demonstrate the efficacy of our method.

math.NA↗

arXiv2023

Multi-modal Visual Understanding with Prompts for Semantic Information Disentanglement of Image

Multi-modal visual understanding of images with prompts involves using various visual and textual cues to enhance the semantic understanding of images. This approach combines both vision and language processing to generate more accurate predictions and recognition of images. By utilizing prompt-based techniques, models can learn to focus on certain features of an image to extract useful information for downstream tasks. Additionally, multi-modal understanding can improve upon single modality models by providing more robust representations of images. Overall, the combination of visual and textual information is a promising area of research for advancing image recognition and understanding. In this paper we will try an amount of prompt design methods and propose a new method for better extraction of semantic information

cs.CV↗

arXiv2023

SportsTrack: An Innovative Method for Tracking Athletes in Sports Scenes

The SportsMOT dataset aims to solve multiple object tracking of athletes in different sports scenes such as basketball or soccer. The dataset is challenging because of the unstable camera view, athletes' complex trajectory, and complicated background. Previous MOT methods can not match enough high-quality tracks of athletes. To pursue higher performance of MOT in sports scenes, we introduce an innovative tracker named SportsTrack, we utilize tracking by detection as our detection paradigm. Then we will introduce a three-stage matching process to solve the motion blur and body overlapping in sports scenes. Meanwhile, we present another innovation point: one-to-many correspondence between detection bboxes and crowded tracks to handle the overlap of athletes' bodies during sports competitions. Compared to other trackers such as BOT-SORT and ByteTrack, We carefully restored edge-lost tracks that were ignored by other trackers. Finally, we reached the SOTA result in the SportsMOT dataset.

cs.CV↗

arXiv2022

Talent Recommendation on LinkedIn User Profiles

With the increasing amount of information on the Internet, recommender systems are becoming increasingly crucial in supporting people to find and explore relevant content. This is also true in the online recruitment space, with websites such as LinkedIn, Indeed.com, and Monster.com all using recommender systems. In online recruitment, it can often be challenging for companies to find suitable candidates with appropriate skills because of the huge volume of user profiles available. Identifying users which satisfy a range of different employer needs is also a difficult task. Thus, effective matching of user-profiles and jobs is becoming crucial for companies. This research project applies a wide range of recommendation techniques to the task of user profile recommendation. Extensive experiments are conducted on a large-scale real-world LinkedIn dataset to evaluate their performance, with the aim of identifying the most suitable approach in this particular recommendation scenario.

cs.IR↗

arXiv2022

A splitting Hamiltonian Monte Carlo method for efficient sampling

We propose a splitting Hamiltonian Monte Carlo (SHMC) algorithm, which can be computationally efficient when combined with the random mini-batch strategy. By splitting the potential energy into numerically nonstiff and stiff parts, one makes a proposal using the nonstiff part of $U$, followed by a Metropolis rejection step using the stiff part that is often easy to compute. The splitting allows efficient sampling from systems with singular potentials (or distributions with degenerate points) and/or with multiple potential barriers. In our SHMC algorithm, the proposal only based on the nonstiff part in the splitting is generated by the Hamiltonian dynamics, which can be potentially more efficient than the overdamped Langevin dynamics. We also use random batch strategies to reduce the computational cost to $\mathcal{O}(1)$ per time step in generating the proposals for problems arising from many-body systems and Bayesian inference, and prove that the errors of the Hamiltonian induced by the random batch approximation is $\mathcal{O}(\sqrt{Δt})$ in the strong and $\mathcal{O}(Δt)$ in the weak sense, where $Δt$ is the time step. Numerical experiments are conducted to verify the theoretical results and the computational efficiency of the proposed algorithms in practice.

math.NA↗