Stay Hungry | Stay Foolish
Hello. I'm Runyu Guan, an SDE in Amazon.com working on AWS issue management. I graduated from University of Southern California, also a Research Assistant in Spatial Computing Lab (moved to University of Minnesota, Twin Cities), ICAROS Lab and IHP Lab (USC ICT) at USC.
As a researcher, I have worked on arbitrary-shape text detection and recognition on historical maps, extracting human preference in industrial assembly tasks and interpretable multi-modality deep learning detection for human mental health etc. My research interests include Multi-modality Deep Learning applied in Computer Vision and Natural Language Processing, Knowledge Graph and Explainable Deep Learning.
If you would like to cooperate with me, feel free to send me an email!
—————— 张载 (Zai Zhang from Song dynasty)
21st IEEE International Conference on High Performance Computing and Communications (Proceedings - 21st IEEE International Conference on High Performance Computing and Communications, 17th IEEE International Conference on Smart City and 5th IEEE International Conference on Data Science and Systems, HPCC/SmartCity/DSS 2019, p 208-214, August 2019)
(Tier-1 conference in High Performance Computing and Communications)
GEOAI '21: Proceedings of the 4th ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery, November 2021, Pages 17–26, https://doi.org/10.1145/3486635.3491070
(Top tier conference in GIS and Spatial Data Computing)
17th Annual ACM/IEEE International Conference on Human-Robot Interaction (HRI 2022), Machine Learning in Human-Robot Collaboration: Bridging the Gap Workshop
(Top tier conference in Robotics)
2019 IEEE 35th International Conference on Data Engineering (Proceedings - International Conference on Data Engineering, v 2019-April, p 112-123, April 2019)
(Top tier conference in data and database)
The 14th International Conference on Knowledge Science, Engineering and Management (Knowledge Science, Engineering and Management. KSEM 2021. Lecture Notes in Computer Science, vol 12815. Springer, Cham. https://doi.org/10.1007/978-3-030-82136-4_6)
(Tier-1 conference in Knowledge Science)
The invention relates to a method for querying keywords in a federated RDF database which is feasible and efficient, and the keyword query results in the federated RDF database can be calculated within the acceptable time.
The invention discloses a multi-hop learning method for improving the cooperation level of a multi-agent system, and belongs to the field of artificial intelligence and computer networks. Based on thenetwork evolutionary game theory, an intelligent agent is promoted to adopt a cooperation strategy by enhancing the learning range of the intelligent agent, and the overall income of the system is improved.
A mobile app that connects college teachers and students in terms of one specific research offer.
Also a Research Assistant focusing on Computer Vision, Natural Language Processing, Robotics and Affective Computing
GPA: 4.0; Major GPA: 4.0
National Scholarship (3 times), Best Student of the year, HuaWei Scholarship
American Alliance for International Education
Accomplished project: Robot Prototype for Remote Searching and Rescue
AWS issue management
Research Advisor: Prof. Yaoyi Chiang
Designed an end-to-end deep learning system with Tensorflow and Keras using Convolutional Neural Networks(VGG,ResNet) for arbitrary-shape text detection and recognition on historical maps, accomplished a better performance than Google Vision, PSENet and EAST, cooperating with the Alan Turing Institute who used our model to extract information from historical maps; built our own standardized dataset using StarGAN, CycelGAN and PAL algorithm defeating ICDAR dataset on historical maps, having a paper accepted by SIGSPATIAL 2021.
Programmed algorithms with PyTorch for multi-modality fine-grained geographic entity typing based on unsupervised or weak supervised learning and Natural Language Processing using advanced BERT.
Utilized React Native, Firebase database and Flask to construct a mobile App to help teachers reward students enhancing enthusiasm for learning.
Built a website and mobile App for movie and TV series sharing based on TMDB data using Angular, Bootstrap, Flask, Android.
Research Advisor: Prof. Stefanos Nikolaidis
Contributed in exploiting worker’s preference to help improve Human-Machine Cooperation; abstracted preference with Inverse Reinforcement Learning. Cooperating with NASA, presented the project demo to former NASA astronaut, Garrett Erin Reisman. We had a paper published in HRI 2022.
Applied RGBD image from Kinect and OpenPose to track human skeleton; programmed object detection based on AprilTag; combined skeleton tracking and object detection to recognize actions; upgraded operation code for robot in ROS system.
Research Advisor: Prof. Mohammad Soleymani
Designed an explainable deep learning multi-modality model to predict depression and PTSD with Tensorflow and Keras based on multi-task learning, exceeding runner-up in AVEC 2019 contest.
Employed OpenFACE, OpenSMILE, BERT to extract features and attention mechanism and SHAP to construct explainable model. Analyzed and visualized data with Statsmodels, Scikit-learn, Seaborn.
Participated in requirements analysis and product development seminar, disigned the website prototype, realized the fornt-end interfaces of the website and some back-end modules, deployed it for Commercial Aircraft Corporation of China Ltd.
Automatic Street Lighting Condition Detection System for Tianjin
Solved some technical challenges about displaying dynamic charts, figured out the data transmission protocal of sensers after many tests, engineered data parser programs, designed and realized the front-end interfaces and back-end modules of the system.
Aerosol Experimental Sensor Management System
Desigend and decorated the user-interface, constrcuted some back-end modules, assembled, deployed and tested the entire system with other engineers.
Using: LabView, Photoshop
Updated Company Website
Updated and repaired the front-end, updated some back-end modules, built, delivered and examined the whole website.
Research Advisor: Prof. David Clark, University of South Florida; International Project Launched by American Alliance for International Education
Robot Prototype for Remote Searching and Rescue
Searched for related resources and conducted a market survey for feasibility and technical solutions. Programmed embedded codes for robots and an mobile control android app. Constructed mechanical components, utilized the 3D printing and fabricated the robot. Operated the whole system.
Using: Arduino, C, Android, Solid Works, Bluetooth
Research Advisor: Prof. Lei Zou, Peking University
Accelerating Partial Evaluation in Distributed SPARQL Query Evaluation
Installed and deployed a distributed system using Spark and on servers with Linux. Optimized experimental codes and Linux parameters, organized experiments. Analyzed the experimental results. Having a paper published in ICDE 2019.
Using: Spark, MapReduce, Linux, Scala, Python
Effective Keyword Search over Federated RDF Systems
Invented an algorithm to reduce searching time on the schema-graph in our federated RDF system. Designed the whole system with my mentor and implemented it with a displaying website. Composed the first-author paper for the algorithm demo. Deployed the system on a server. We had a validated patent for this project.
Research Advisor: Prof. He Zhang, Nanjing University
A Systematic Literature Review of Code-reviewer Recommendation Systems
Summarized and classified methods and algorithms after literature review. Compared their experiments and experimental data. Evaluated the data, drew our unique conclusions, put forward some problems haven’t been solved. Shared our conclusions in a speech.
Using: Latex, Java, Python, JSON
Research Advisor: Prof. Tao Wang, Hunan University
Multi-hop Learning’s Effects over the Rate of Multi-agent Cooperation
Improved codes of the experimental platform. Deliberated the experimental scheme with my mentor and ran experiments. Analyzed and visualized the results. Wrote the first author paper and published in KSEM 2021. Our new mechanism can greatly enhanced cooperation rate.
Using: C++, Python, Latex, Matlab
An Improved Defense Model for Deceptive Attack in Honeypot-Enabled Networks
Put forward the core algorithm based on Bayesian model and Relative Historical Income model. Programmed codes for the experiment platform and conducted experiments. Evaluated and visualized the results. Composed the first author paper, published in HPCC 2019.
Using: Java, Latex, Python, Matlab
A Real-time Tool Recognition System Based on Deep Learning
Created a unique dataset based on specific tools given by the customer. Devised the core algorithm model based on VGG. Engineered the model with an Android app and a server-side system. Planned tests and improved the model parameters.
Using: Vue.js, Pytorch, Python, OpenCV, Android, Java, MySQL, JavaEE
A mobile app that connects college teachers and students in terms of one specific research offer.
Navigated the object-oriented requirements analysis. Confirmed the technical pattern. Desigend and fabricated the database, the android app, the back-end server system and the management website. Dispersed internal test for the app. Communicated with college officials and promoted this app in Hunan University.
Advisor: Prof. Cheng Xv, Hunan University
A multi-dimensional warning system prototype for railway construction team workers, cooperating with Chinese Ministry of Railways.
Operated surveys among railway workers. Identified appropriate sensors after hundreds of practical tests. Invented and built programs in Embedded Microcontroller Units, an Android app, a server-side system and a management website.
《新世纪福音战士：终》在日本上映已经是半年前的事情了，根据日本国内政策，最近登录网络流媒体已经算是很迅速，我也趁势去下了个4K画质版爽了一爽。先说结论，EVA的新剧场版四部曲（序，破，Q，终）私以为是应该放在动漫爱好者一生必看名单中。在EVA自己的系列里，新四部曲画面和制作技术自然是顶配，内容和老TV版以及老剧场版《真心为你》相比，若是前三部则略显平淡，但加上《终》之后可以说各有千秋。如果单说《终》，片如其名，毫无疑问是EVA整个系列最尽善尽美的终局... see more
最近沸沸扬扬的话题很多，其中一个经久不衰并且不断自我复杂化的话题就是“XXX是不是媚男”，这里的XXX可以放入“人”比如鞠婧祎、半藏森林，也可以放入“物”比如JK、lo裙，也可以放入“概念”比如婚姻、彩礼... see more
不要忘了我。“她说。 “永远不忘记你。“我说。 “也许以后没机会再见到你了，不过，无论丢到那里，我都会永远记得你和直子。” 我看看她的眼睛，她哭了。我禁不住吻了她。虽然周围经过的人频频盯看我们看，但我已经不在意了。我们活看，只须考虑怎样活下去就够了。 “祝你幸福。“分手之际，玲子对我说。“我能向你忠告的全都说完了，再也没什么好说的，只能祝你幸福。让我和直子那一份的幸福都给予你。” 我们握手告别。
我打电话给阿绿，说无论如何都要跟她谈一谈。我说我有很多话要说，必须对她说。在这个世界上，除她以外别无所求。我想见她，一切的一切从头开始来过。 阿绿在电话的另一端，沉默了好久。彷佛全世界的细雨下在全世界的青草地上似的，沉默无声。那段时间，我闭起眼睛，额头一直压在玻璃窗上，终于阿绿开口了。她用平静的声音说：“现在你在哪里？” 我现在在哪里？ 我继续握住听筒抬起脸来，看看电话亭的四周。如今我在什么地方？我不知道那是什么地方。我猜不看。到底这里是那里？映入我眼帘的只是不知何处去的人们，行色匆匆地从我身边走过去。而我只能站在那个不知名的地方，不停地呼唤阿绿的名字。
In recent years, there is always a controversial issue that movies and television are bad for children and teenagers. From my perspective, I disagree with that statement.
To begin with, movies and television are new types of art which means that they can give children and teenagers a new way to feel the beauty of the life and the world. Many experts and artists agree that movie is the 7th art, that is why there are many film festivals to evaluate movies and find the best movie of the year. For children and teenagers, movies and television give them a vivid way to figure out the beauty of the world. For example, when I was a child, I could not know what did Paris look like just from many books. On the contrary, after I saw a movie made in Paris, I could imagine many images about Paris in my mind and when I read books about Paris again, that movie can help me understand those books better.
Additionally, movies and televisions can help young people understand many problems vividly and make them do something to solve them. When young people read environmental problems from newspapers, it's difficult for them to imagine how bad that problem will be. For example, when talking about global warming, many people could not believe that a warmer earth will lead to lots of disasters, but when watching a television program about global warming which shows that polar bears have no where to live because of less ice, most people will agree that global warming is a serious problem. That is the power of movies and television. On top of that, as young people have less living experience vivid descriptions are more important to them than to grown-ups.
Admittedly, watching too much movies and television will make young people become sofa potatoes. But, from my perspective, the government and schools can do many things to protect children and teenagers from watching too much movies and television. Besides, nowadays there are many interesting things other than movies and television such as sports and travelling which are widely introduced by advertisements. So, less and less children will think that movies and television are the only way they can relax themselves. As a result, I believe that with the help of governments and schools, parents can make their children and teenagers to stop watching too much movies and television and pay attention to their body health.
All in all, as the reason mentioned above, I disagree with the argument that movies and television are bad for children and teenagers.
想起这个话题是因为前两天看到一个朋友发的段子，说一位体制内的科研大牛对下属说了一句：“我读过的书堆起来可以把你们火化了。”轻描淡写中霸气侧漏，让我身不能至心向往之。古有军阀，今有财阀，这些大牛们用“学阀”来称呼似很贴切... see more
这两天赛博考古，把《FSN:HF》没看的二和三翻出来看。其实我对FSN的IP从来是有就看，追新番，电影院支持，但因为第一部给我的观感实在不咋地，加上因为这部电影把一个要好的朋友气哭了，导致我在五味杂陈之下没及时看后面的二和三。昨天重看之后，萌生了一些奇奇怪怪的感触，正好我从来没有给FSN写过什么杂感，于是在这里扯一扯... see more