Yingzhen Li

Yingzhen Li

Auto-generated profile, based on publicly available information

Yingzhen Li

Imperial College London

Research Focus & Metrics

Academic Career Length

14Years (2013-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications

1 July 2014
Verified
Conference Paper
Closed access
HMHao Meng
Hao Meng
YLYingzhen Li
Yingzhen Li
For the uncertain impact of ship air wake and deck motion to the landing process, the traditional automatic carrier landing system lacks flexibility and accuracy. This paper designs fuzzy controller for automatic carrier landing of aircraft. The application examined is a F/A-18 Fuzzy Logic Automatic Carrier Landing System (FACLS). The FACLS has seven inputs, two outputs, and several hundred fuzzy rules. The fuzzy rules are distributed among four rule bases that are embedded in a classical control structure. The use of fuzzy controller was different from many applications since fuzzy logic is used not only to model a set of human rules, but to combine the best features of human and automatic control approaches that cannot be easily accommodated in conventional control systems. The FACLS is tested in simulation and compared with the conventional F/A-18 Automatic Carrier Landing System. Results suggest that fuzzy controller could reduce the aircraft's landing error and improve the landing quality.


31 March 2024 in arXiv
Verified
Preprint
Open access
PDF file
HYHee Suk Yoon
Hee Suk Yoon
EYEunseop Yoon
Eunseop Yoon
JTJoshua Tian Jin Tee
Joshua Tian Jin Tee
MHMark Hasegawa-Johnson
Mark Hasegawa-Johnson
YLYingzhen Li
Yingzhen Li
CYChang D. Yoo
Chang D. Yoo
In deep learning, test-time adaptation has gained attention as a method for model fine-tuning without the need for labeled data. A prime exemplification is the recently proposed test-time prompt tuning for large-scale vision-language models such as CLIP. Unfortunately, these prompts have been mainly developed to improve accuracy, overlooking the importance of calibration, which is a crucial aspect for quantifying prediction uncertainty. However, traditional calibration methods rely on substantial amounts of labeled data, making them impractical for test-time scenarios. To this end, this paper explores calibration during test-time prompt tuning by leveraging the inherent properties of CLIP. Through a series of observations, we find that the prompt choice significantly affects the calibration in CLIP, where the prompts leading to higher text feature dispersion result in better-calibrated predictions. Introducing the Average Text Feature Dispersion (ATFD), we establish its relationship with calibration error and present a novel method, Calibrated Test-time Prompt Tuning (C-TPT), for optimizing prompts during test-time with enhanced calibration. Through extensive experiments on different CLIP architectures and datasets, we show that C-TPT can effectively improve the calibration of test-time prompt tuning without needing labeled data. The code is publicly accessible at https://github.com/hee-suk-yoon/C-TPT.

1 January 2014 in SSRN Electronic Journal
Verified
Preprint
Open access
PDF file
TCThomas J. Chemmanur
Thomas J. Chemmanur
YLYingzhen Li
Yingzhen Li
JXJing Xie
Jing Xie
AZAnthony Zhu
Anthony Zhu
We propose a “noisy signaling" hypothesis of open market share repurchase (OMSR) programs, where the equity market equilibrium that prevails after OMSR program announcements is a partial pooling rather than a fully separating equilibrium. We argue that two complementary mechanisms, namely, actual share repurchases by firms and information production by institutions, serve to reduce the residual equity market information asymmetry facing firms subsequent to OMSR program announcements. We test the implications of our noisy signaling hypothesis using transaction-level data on trading by institutions and by a subsample of identified hedge funds, and find strong support for the above hypothesis.

29 October 2025 in arXiv
Verified
Preprint
Open access
PDF file
NKNaoki Kiyohara
Naoki Kiyohara
EJEdward Johns
Edward Johns
YLYingzhen Li
Yingzhen Li
Stochastic differential equations (SDEs) are well suited to modelling noisy and irregularly sampled time series found in finance, physics, and machine learning. Traditional approaches require costly numerical solvers to sample between arbitrary time points. We introduce Neural Stochastic Flows (NSFs) and their latent variants, which directly learn (latent) SDE transition laws using conditional normalising flows with architectural constraints that preserve properties inherited from stochastic flows. This enables one-shot sampling between arbitrary states and yields up to two orders of magnitude speed-ups at large time gaps. Experiments on synthetic SDE simulations and on real-world tracking and video data show that NSFs maintain distributional accuracy comparable to numerical approaches while dramatically reducing computation for arbitrary time-point sampling.