About

I’m a grad student at New York University and a member of CILVR Group. My research interest lies in deep reinforcment learning, unsupervised learning and natural language understanding.

Prior to NYU, I completed my bachelors at IIT, Mandi during which I also spent an year at Technical University, Munich (TUM).

Projects & Publications

• Learning When to Communicate
  Amanpreet Singh^*, Tushar Jain^*, Sainbayar Sukhbaatar
  ICLR 2019
  [Paper] [Code]

• Crosslingual Transfer for Inflection Generation
  Tushar Jain, Katharina Kann
  Research on morphological generation in low resource cross-lingual settings. WIP

• ---

  Tushar Jain, Shubham Chandel, Mikael Henaff, Yann LeCun
  Using model-based reinforcement learning in text-based games. WIP

• Question Generation
  Rajat Agarwal, Tushar Jain, Kumar Mehta
  Survey research project on question generation via machine comprehension and achieved SOTA results for the task using attention with seq-to-seq with beam search.
  [Report]

• Generative Adversarial Imitation Learning
  Under Prof. Rob Fergus
  Research on imitation learning using generative model. Designed a novel method for using GANs as environment simulator for model-based Reinforcement Learning.
  

• Genomics Trait & Disease Prediction
  Under Prof. Lawrence Sirovich
  Research on genomics data (GWAS), where #input-features $\gg$ #training-examples, making it difficult to avoid overfitting, for disease prediction on SNPs.
  

• Biometric Data Synthesis Using GAN
  Under Prof. Aditya Nigam
  Research on using multiple generators and discriminators to speed-up GAN training. Multiple discriminators with varying architecture provided empirical speedup during training time.