I am a undergraduate researcher under the advisement of professor Samuel Bowman at NYU with an interest in machine learning, especially in the joint field of natural language understanding and deep learning. I am affliated to Machine Learning for Language group, which is part of larger CILVR lab founded by Professor Yann LeCun.
- I will Join Google Research as a intern in 2018 summer.
- I build a very strong agent for Pommerman. In FFA mode, the winning chance against Simple Agent baseline is 68%, as a comparison, the expected winning chance is 25% if our agent is same as the baseline. Will I win the competition in NIPS? Let’s see.
- I received student travel award from ICLR2018. Thank you ICLR!
- I broke the record on MultiNLI dataset during summer. The paper Natural Language Inference over Interaction Space is out. Our approach reduced 20% of error rate against the strongest baseline at the time. It is accepted at ICLR 2018 as a poster paper.
- My preprint manuscript of Ruminating Reader: Reasoning with Gated Multi-Hop Attention is out. Ruminating Reader extends BiDAF to multi-hop attention version and improves performance by approximately 2% on SQuAD.
- Our NYUDaD (NYU Data scientists as Doctors) team won the 23rd place in 2017 Data Science Bowl. The competition requires candidates to use computer vision to identify malicious lung cancer nodule.
- I have solved the Ms. Pac-Man Game (v6.2 simulator) with weighted score function and search strategy. My agent triples the best score from 2011 competition and touches the walltime without running out of lives.
- I have solved text game Machine of Death via deep reinforcement learning, specifically, multi-step Q-learning. The game is stocastic and not solved in paper He et al.