Research Intern
Uber ATG
Oct 17 - Present
Contract Developer
Google Summer of Code
May 16 - Aug 16
Research Intern
Virginia Tech
May 17 - Aug 17
B.Sc. in Computer Science
National Tsing Hua University
Sep 12 - Jun 16
Bio. I am currently an intern at Uber Advanced Technologies Group, where I work on machine learning for autonomous driving perception and control, advised by Prof. Raquel Urtasun. In addition to my research experience, I build interesting open-source projects such as DeepLearningFlappyBird, which has been featured on Hacker News.

Previously, I am lucky to have the opportunities to work with Prof. Min Sun at National Tsing Hua University, Dr. Ming-Yu Liu at NVIDIA, and Prof. Jia-Bin Huang at Virginia Tech.

See my CV for more details.


Deep 360 Pilot: Learning a Deep Agent for Piloting through 360° Sports Videos
Watching a 360° sports video requires a viewer to continuously select a viewing angle, either through a sequence of mouse clicks or head movements. To relieve the viewer from this “360 piloting” task, we propose “deep 360 pilot” – a deep learning-based agent for piloting through 360° sports videos automatically. At each frame, the agent observes a panoramic image and has the knowledge of previously selected viewing angles. The task is to shift the current viewing angle (i.e. action) to the next preferred one (i.e., goal).
(* indicates equal contribution)
Yen-Chen Lin*, Hou-Ning Hu*, Ming-Yu Liu, Hsien-Tzu Cheng, Yung-Ju Chang, Min Sun
CVPR 2017 (Oral)
Tactics for Adversarial Attack on Deep Reinforcement Learning Agents
We introduce two tactics to attack agents trained by deep reinforcement learning algorithms using adversarial examples: Strategically-timed attack: the adversary aims at minimizing the agent's reward by only attacking the agent at a small subset of time steps in an episode. Enchanting attack: the adversary aims at luring the agent to a designated target state.
Yen-Chen Lin, Zhang-Wei Hong, Yuan-Hong Liao, Meng-Li Shih, Ming-Yu Liu, Min Sun
IJCAI 2017
ICLR 2017 workshop
Detecting Adversarial Attacks on Neural Network Policies with Visual Foresight
In this paper, we propose a defense mechanism to defend reinforcement learning agents from adversarial attacks by leveraging an action-conditioned frame prediction module. Our core idea is that the adversarial examples targeting at a neural network-based policy are not effective for the frame prediction model. By comparing the action distribution produced by a policy from processing the current observed frame to the action distribution produced by the same policy from processing the predicted frame from the action-conditioned frame prediction module, we can detect the presence of adversarial examples.
Yen-Chen Lin, Ming-Yu Liu, Min Sun, Jia-Bin Huang
NIPS 2017 workshop
Tell Me Where to Look: Investigating Ways for Assisting Focus in 360° Video
One challenge of watching 360° videos is continuously focusing and re-focusing intended targets. To address this challenge, we developed two Focus Assistance techniques: Auto Pilot, and Visual Guidance. We conducted an experiment to compare users’ video-watching experience and sickness using these techniques, and obtained their qualitative feedback. Our results provide design implications for better 360° video focus assistance.
Yen-Chen Lin, Yung-Ju Chang, Hou-Ning Hu, Hsien-Tzu Cheng, Chi-Wen Huang, Min Sun
CHI 2017
Semantic Highlight Retrieval
Finding highlights relevant to a text query in unedited videos has become increasingly important due to their unprecedented growth. We refer this task as semantic highlight retrieval and propose a query-dependent video representation for retrieving a variety of highlights. Our method consist of two parts: (1) “viralets”, a mid-level representation bridging between visual and semantic spaces; (2) a novel Semantically-Modulation (SM) procedure to make viralets query-dependent (referred to as SM viralets).
Kuo-Hao Zeng, Yen-Chen Lin, Ali Farhadi, Min Sun
ICIP 2016

Side Projects

DeepLearningFlappyBird is a Flappy Bird hack using Deep Reinforcement Learning. This project follows the description of the Deep Q Learning algorithm described in Playing Atari with Deep Reinforcement Learning and shows that this learning algorithm can be further generalized to the notorious Flappy Bird. See the demo video here.
awesome-watchos is a curated list of awesome watchOS frameworks, libraries, sample apps. It is created with the purpose of helping people to get the hang of apple watch programming.

Work Experiences

Google Summer of Code
In 2016, I participated Google Summer of Code program and worked with scikit-learn community. My mission is to make implementation of many algorithms, such as Stochastic Gradient Descent, Coordinate Descent, support Cython fused types and therefore reduce memory waste.
Real Forest
Real Forest is a feature I developed for well-known iOS app - Forest: Stay focused, be present when I was an intern there. This feature let user plant a real tree in the world using the coins they earned in the app. So far, it has helped planting 8000+ real trees in India and Zambia.