Fan-Yun Sun
I am seeking for a Ph.D. position that starts from 2020 Fall.
sunfanyun [at]
Another me.
I love movies and sports, especially basketball and table tennis.


Fan-Yun Sun obtained his bachelor's degree in Computer Science and Information Engineering at National Taiwan University, where he conducted research at Machine Discovery and Social Network Mining Laboratory, advised by Prof. Shou-De Lin. Subsequently, he worked with Prof. Jian Tang at MILA and Prof. Jure Leskovec at Stanford. If you would like to learn more about me, please see my [ CV ] or contact me at sunfanyun [at]


  • Machine Learning / Deep Learning
  • Structure Learning / ML on Graphs
  • Reinforcement Learning / Multiagent Systems


National Taiwan University (NTU)
B.S. in Computer Science, 2019
  • Overall Class Rank: 3 / 123
  • Overall GPA: 4.20 / 4.30
  • CS Major GPA: 4.28 / 4.30

Timeline & Experiences

2019 July - 2019 Oct.
Visiting Student Researcher @ Stanford
Conducted data mining research on temporal graph dataset.
Supervisor: Prof. Jure Leskovec
2019 Jan. - 2019 May
Research Intern @ MILA
First authored two papers on graph learning.
See publications.
Supervisor: Prof. Jian Tang
2018 Mar. - 2018 Sep.
ML Engineer Intern @ Appier
Researched and implemented
- RNN-based recommendation methods
- graph-based recommendation methods
Supervisor: Prof. Hsuan-tien Lin
2018 Jan. - 2018 Feb.
Research intern @ WorldQuant
Conduct quantitative research (finding alphas).
Gold level on Websim.
2017 - 2018
Microsoft Student Partner @ Microsoft
Workshop lecturer.
2017 July - 2017 Aug.
Software Engineering Intern @ Google
Writing open source tools for Android Taipei Team.
A tool for labeling dependencies with a web interface [ code ]
A Java Load-Libraries analyzer using framework Soot [ code ]
Supervisor: Logan Chien, Hung-ying Tyan
2015 - 2019
Undergraduate student & researcher @ NTU
Computer Science department
Machine Discovery and Social Network Mining Laboratory & Multimedia indexing, Retrieval, and Analysis Lab
Primary focus:
- (Multiagent) Reinforcement learning
- Graph representation Learning
- Time series prediction
- Instance Segmentation on medical image
He was Born


InfoGraph: Unsupervised and Semi-supervised Graph-Level Representation Learning via Mutual Information Maximization
This paper studies learning the representations of whole graphs in both unsupervised and semi-supervised scenarios using mutual information maximization.
ICLR 2020 (Spotlight) Addis Ababa, Ethiopia
vGraph: A Generative Model for Joint Community Detection and Node Representation Learning
In current literature, community detection and node representation learning are usually independently studied while they are actually highly correlated. We propose a probabilistic generative model called vGraph to learn community membership and node representation collaboratively. We also show that the framework of vGraph is quite flexible and can be easily extended to detect hierarchical communities.
NeurIPS, 2019 Vancouver, BC
A Regulation Enforcement Solution for Multi-agent Reinforcement Learning
In this paper, we proposed a framework to solve the following problem: In a decentralized environment, given that not all agents are compliant to regulations at first, can we develop a mechanism such that it is in the self-interest of non-compliant agents to comply after all. We utilized empirical game-theoretic analysis to justify our method.
AAMAS, 2019 Montreal, QC
Designing Non-greedy Reinforcement Learning Agents with Diminishing Reward Shaping
This paper intends to address an issue in multi-agent RL that when agents possessing varying capabilities. We introduce a simple method to train non-greedy agents with nearly no extra cost. Our model can achieve the following goals: non-homogeneous equality, only need local information, cost-effective, generalizable and configurable.
AAAI/ACM conference on AI, Ethics, Society 2018 (Oral) New Orleans, LA
A Memory-Network Based Solution for Multivariate Time-Series Forecasting
Inspired by Memory Network for solving the question-answering tasks, we proposed a deep learning based model named Memory Time-series network (MTNet) for time series forecasting. Additionally, the attention mechanism designed enable MTNet to be interpretable.
ANS: Adaptive Network Scaling for Deep Rectifier Reinforcement Learning Models
This work provides a thorough study on how reward scaling can affect performance of deep reinforcement learning agents. We also propose an Adaptive Network Scaling framework to find a suitable scale of the rewards during learning for better performance. We conducted empirical studies to justify the solution.
Organ At Risk Segmentation with Multiple Modality
In real world scenario, doctors often utilize multiple modalities. In this paper, we propose to use Generative Adversarial Network to perform CT to MR transformation to synthesize MR images instead of aligning two modalities. The synthesized MR can be jointly trained with CT to achieve better performance.

Teaching & Talks

Teaching Assistant

    Data Structures and Algorithms (Spring 2017)
  • Designe algorithm problems, provide test cases, judge programs and solutions.
  • Correct students' homework and provide guidance at TA time.


Honors & Awards

    Machine Learning / Deep Learning

  • Ranked 19th (out of 4180) / KDD CUP - Main Track / 2018
  • Ranked 4th (out of 4180) / KDD CUP - Specialized Prize for long term prediction / 2018
  • Top 12 (out of 149) / Formosa Speech Grand Challenge - Talk to AI (Warm-Up) / 2017 [ code ]
  • Travel Grant for NeurIPS 2019 / Appier / 2019
  • Publication Grant for AAMAS 2019 / National Taiwan University (MSLAB) / 2019
  • Travel Grant for AIES 2018 / National Taiwan University (MSLAB) / 2018
  • ML Research Project Grant / Institute for Information Industry of Taiwan / 2018

    Competitive Programming

  • Top 1000 / Google Code Jam / 2017 [ code ]
  • Top 1000 / Facebook Hacker Cup / 2017
  • 1st Place / ACM-ICPC 2016 Asia-Manila Regionals / 2016 [ image] [ code ]
  • 2nd Place / Newcomers for ACM-ICPC Taiwan Online Programming Contest / 2016
  • 3rd Place / NTU ACM ICPC Ranking / 2016


  • Microsoft Student Partner of the year / Microsoft / 2018
  • Presendential Award (2 times) / National Taiwan University / 2015 Fall, 2016 Spring
  • Best project award (out of 40+ groups) / Probability Course Final Project Contest / 2017
  • 3rd place (out of 280+ students) / Data Structure and Algorithm Final Project / 2016 [ code ]
  • Finalist (Top 30) / International Physics Olympiad Domestic Final / 2014

Selected Other Projects

Neural Network as Neural Network Input
In this paper, we extended the realm of graph benchmark dataset to computation graphs of neural networks. We web scraped tensorflow model files using github API and annotated them with associated meta data, and we benchmarked them using existing graph classification methods.
[ pdf ]
Intelligent Conversational Bot of TV / Movie (2017)
Designed and implemented an AI chatbot of TV / Movie. Involved in crawling data, training RNN-NLU for language understanding, experimenting RL-based dialogue tracker and seq2seq model for natural language generation.
Formosa Speech Grand Challenge-Talk to AI Warm-Up Match (2017)
Top 12 solution of a response selection chatbot competition held by ministry of science and technology of Taiwan. Our solution is an ensemble of the following three models: Sequential Matching Network (left figure), a RNN model, and simple averaging of word vectors.
[ code ]
Collaborative Editing of Vim Enhanced (2018)
Enhanced CoVim by eliminating the need of public IPs or a central server(utilize TCP Hole Punching).
E-monitor (2017)
MAIN IDEA: An android app integrated with IOT device to monitor electronic devices in real-time. To register a electronic device, simply scan QR code attached to the device. This project include hardward realization.
Pike (2015)
Designed an app to help students park their bikes easily and efficiently from scratch. Our team went through process of needfinding, low-fi, mid-fi, high-fi prototyping to user interface implementation in Java (Android).