Tianyang Han

ORCID: 0000-0001-5541-0806
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Traffic Prediction and Management Techniques
  • Human Mobility and Location-Based Analysis
  • Traffic control and management
  • Natural Language Processing Techniques
  • Transportation Planning and Optimization
  • Distributed and Parallel Computing Systems
  • Data Management and Algorithms
  • Topic Modeling
  • Autonomous Vehicle Technology and Safety
  • Reinforcement Learning in Robotics
  • Multimodal Machine Learning Applications
  • Scheduling and Optimization Algorithms
  • Urban Transport and Accessibility
  • Robotic Path Planning Algorithms
  • Transportation Safety and Impact Analysis
  • Rough Sets and Fuzzy Logic
  • Interpreting and Communication in Healthcare
  • Artificial Intelligence in Law
  • Tensor decomposition and applications
  • Graph Theory and Algorithms
  • Single-cell and spatial transcriptomics
  • Elevator Systems and Control
  • Power Line Inspection Robots
  • Gene expression and cancer classification
  • Advanced Graph Neural Networks

Hong Kong Polytechnic University
2024

The University of Tokyo
2019-2024

Shenzhen University
2022

10.18653/v1/2024.emnlp-main.895 article EN Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing 2024-01-01

Multimodal Large Language Models (MLLMs) excel in generating responses based on visual inputs. However, they often suffer from a bias towards similar to their pretraining corpus, overshadowing the importance of information. We treat this as "preference" for statistics, which hinders model's grounding input. To mitigate issue, we propose Bootstrapped Preference Optimization (BPO), conducts preference learning with datasets containing negative bootstrapped model itself. Specifically, following...

10.48550/arxiv.2403.08730 preprint EN arXiv (Cornell University) 2024-03-13

Advances in single-cell RNA sequencing (scRNA-seq) technologies has provided an unprecedent opportunity for cell-type identification. As clustering is effective strategy towards identification, various computational approaches have been proposed scRNA-seq data. Recently, with the emergence of cellular indexing transcriptomes and epitopes by (CITE-seq), cell surface expression specific proteins on same can be captured, which provides more comprehensive information analysis. However, existing...

10.1093/bib/bbac347 article EN Briefings in Bioinformatics 2022-08-31

The deployment of multimodal large language models (MLLMs) has brought forth a unique vulnerability: susceptibility to malicious attacks through visual inputs. We delve into the novel challenge defending MLLMs against such attacks. discovered that images act as "foreign language" is not considered during alignment, which can make prone producing harmful responses. Unfortunately, unlike discrete tokens in text-based LLMs, continuous nature image signals presents significant alignment...

10.48550/arxiv.2401.02906 preprint EN other-oa arXiv (Cornell University) 2024-01-01

Incomplete data is common and unavoidable in the data-driven intelligence transportation system. There are several studies on traffic imputation, while how to utilize temporal-spatial information still under discussion. In this paper, we propose an imputation algorithm based matrix completion graphs. To derive appropriate estimation matrix, temporal spatial dependencies considered as The smoothness of graphs added objective function regularization terms. Then, a heuristic method proposed...

10.1109/itsc.2019.8917365 article EN 2019-10-01

Prediction is an important part of the traffic management system (TMS) which supports route planning, dynamic control, and information provision. We developed a multi-dimensional learning machine for predicting speed. Proposed methodology considered both historical experience near past observation data by combining convolutional neural network (CNN) with tensor decomposition (TD) predictor. TD based method treated as effective predictor considering temporal-spatial neighbourhood data....

10.1016/j.trpro.2020.08.125 article EN Transportation research procedia 2020-01-01

Trip purpose plays a critical role in reflecting human mobility behavior. However, it is relatively difficult to determine. With the rapid growth of urban and big mobile data, utilizing these data for trip classification has been long-term objective enhance travel demand behavior models used planning. Although studies on this topic have extensively conducted, most past research preferred relying traveler attributes or histories achieve accurate results. These could be privacy sensitive often...

10.1145/3557915.3560969 article EN Proceedings of the 30th International Conference on Advances in Geographic Information Systems 2022-11-01

Trip purpose is essential information supporting tasks in intelligent transportation systems, such as travel behaviour comprehension, location-based service, and urban planning. The observation of trip a necessary aspect surveys. However, owing to the sampling volume, survey budget, frequency, relying solely on surveys era big data difficult task. There has long been demand for an accurate, generalizable, robust inference method purposes. Although existing studies contributed significant...

10.1109/tits.2022.3213969 article EN IEEE Transactions on Intelligent Transportation Systems 2022-11-03

10.1109/tits.2024.3405171 article EN IEEE Transactions on Intelligent Transportation Systems 2024-01-01

The deployment of multimodal large language models (MLLMs) has demonstrated remarkable success in engaging conversations involving visual inputs, thanks to the superior power (LLMs). Those MLLMs are typically built based on LLMs, with an image encoder process images into token embedding space LLMs. However, integration modality introduced a unique vulnerability: MLLM becomes susceptible malicious inputs and prone generating sensitive or harmful responses, even though LLM been trained textual...

10.48550/arxiv.2409.11365 preprint EN arXiv (Cornell University) 2024-09-17

10.1109/itsc58415.2024.10920107 article EN 2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC) 2024-09-24

<div>Trip purpose is essential information supporting many downstream tasks in intelligent transportation systems, such as travel behaviour comprehension, location-based service, and urban planning. The observation of trip a necessary aspect surveys, it difficult to obtain clear annotated purposes by another approach. However, the limitations sampling volume, survey budget, frequency make rely solely on surveys era big data. There has long been demand for methods accurately infer...

10.36227/techrxiv.19322279.v1 preprint EN cc-by 2022-03-11

<p>Trip purpose is essential information supporting tasks in intelligent transportation systems, such as travel behaviour comprehension, location-based service, and urban planning. The observation of trip a necessary aspect surveys. However, owing to the sampling volume, survey budget, frequency, relying solely on surveys era big data difficult task. There has long been demand for an accurate, generalizable, robust inference method purposes. Although existing studies contributed...

10.36227/techrxiv.19322279 preprint EN cc-by 2022-03-11

<p> </p> <p>Trip purpose is essential information supporting tasks in intelligent transportation systems, such as travel behaviour comprehension, location-based service, and urban planning. The observation of trip a necessary aspect surveys. However, owing to the sampling volume, survey budget, frequency, relying solely on surveys era big data difficult task. There has long been demand for an accurate, generalizable, robust inference method purposes. Although existing...

10.36227/techrxiv.19322279.v2 preprint EN cc-by 2022-08-08

<p>Trip purpose is essential information supporting tasks in intelligent transportation systems, such as travel behaviour comprehension, location-based service, and urban planning. The observation of trip a necessary aspect surveys. However, owing to the sampling volume, survey budget, frequency, relying solely on surveys era big data difficult task. There has long been demand for an accurate, generalizable, robust inference method purposes. Although existing studies contributed...

10.36227/techrxiv.19322279.v3 preprint EN cc-by 2022-11-01

With the development of detection and computation techniques, use reinforcement learning (RL) in traffic signal control problems is widely discussed. After formulating diverse isolated RL agents to one intersection design, most existing studies tend directly duplicate for large-scale coordinated problems. However, two questionable challenges are 1) commonly differs from owing different objectives; 2) coordination necessity varies under demand. Thus, a naive duplication or aggregation seems...

10.1109/itsc55140.2022.9922269 article EN 2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC) 2022-10-08

Reinforcement learning (RL) methods have been used in traffic signal control for decades. Traditional RL controller with model-free design treat states as a Markov Process (MP) to approximate future benefit of strategy and improve its policy discretized action space. In such treatment, the statistical connection between state produced might mismatch theoretical understanding. The mismatching can lead invalid when under differing from assumption. To enhance inference processes engineering...

10.1109/itsc55140.2022.9922584 article EN 2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC) 2022-10-08
Coming Soon ...