About Me
I am a research engineer in the Search Recommendation Center of Social Platform and Application Line at Tencent since 2021. My work mainly focuses on optimizing deep models for short video recommendations in QQ, using large-scale user behavior data.
Before joining Tencent, I received my Master's degree from the School of Computer Science and Technology, Harbin Institute of Technology, Shenzhen in 2021 where I was involved in the research of computer vision applied on remote sensing images.
Prior to that, I received my Bachelor's degree from the School of Computer and Information Engineering, Henan University in 2018 where I got the qualification for the recommended postgraduate with a GPA of 3.66/4 (rank 2/87).
Currently, I have got some English language grades, such as the IELTS(overall score of 6.0 and specifically, writing:6.5, speaking:6.0, reading:5.5, listening:5.5 ) and CET-6(487 marks).
Experience
Tencent
Recommended Algorithm Engineer
2021-Now
-
Responsible for optimizing the short video recommendation model of QQ Kandian. Built a personalized recommendation service for QQ Kandian Friend Channel from 0 to 1, which significantly improved the users' experience and increased the viewing time of users by 25.46% and the interaction rate by 18.06%.
-
Through optimizing the recommendation model on features and structures, which brought a 1.88% improvement in total users' viewing time on QQ Small World. In particular, a list-wise-based model with multi-targets was successfully applied in rerank stage and an expert network with long session modeling and attention mechanism introduced into a MMOE model also improved the recommendation effect. In addition, a clustering network based on attribute features is introduced to enhance the embedding representations of the cold start item ID and user ID.
-
Responsible for building e-commerce personalized recommendation service in QQ Small World, including recall stage, pre-rank stage and rank stage, which doubled the Click-through Rate(CTR) and the Gross Merchandise Volume(GMV) of business cards.
Publication
GCDB-UNet: A Novel Robust Cloud Detection Approach for Remote Sensing Images
Xian Li, Xiaofei Yang, Xutao Li, Shijian Lu, Yunming Ye, Yifang Ban
Knowledge-Based Systems, 2022
[Paper]
[BIbTex]
Introduction: This article mainly studies the problem of semantic segmentation of cloud pixels and non-cloud pixels in satellite remote sensing images. In view of the difficulty in detecting thin cloud areas, this article proposes a Global Context Dense Block(GCDB) based on the Non-local self-attention module and the channel attention module for feature extraction, significantly improving the detection accuracy of thin cloud areas
Patents
Remote sensing image cloud detection method, terminal and storage medium based on fully convolutional network.
Yunming Ye, Xutao Li, Xian Li
Application No. CN202010440430.0
Selected Awards
National Encouragement Scholarship, 2016 and 2017.
First Class Scholarship of Harbin Institute of Technology, 2018.
Outstanding League member-Diligence and Truth-seeking Award of Harbin Institute of Technology, 2019.
Second Class Scholarship of Harbin Institute of Technology, 2019 and 2020.
Second Prize for Technical Contribution of Big Data Technology Research Center, Harbin Institute of Technology, 2019.
Outstanding Graduates of Henan Province, 2018
National Second Prize of Lanqiao Cup (Programming Competition), 2017.
National Third Prize of MathorCup College Mathematical Modeling Challenge, 2017.
National Second Prize of May Day Mathematical Contest in Modeling, 2017.
Excellent Project of College Students' innovation and Entrepreneurship Competition, 2017.
Miscellaneous
I choose to apply for a Ph.D. because I discover my passion to scientific research
during my work at Tencent
and know what social value I want to achieve
and what kind of future I want. So, I intend to do one thing that I love without distractions.
My choice to apply for a Ph.D. may adapt to changes and act accordingly.
I like sports, especially swimming and table tennis.