• Yann Dauphin

    Research Scientist

    @ Google AI



    Publications: Google Scholar

    Email: yann@dauphin.io

    CV: PDF



  • About

    My research focuses on understanding and developing deep learning algorithms. I'm interested in creating deep learning algorithms that can learn with little supervision and to understand the principles of learning.


    I moved in 2019 to Google AI. Before that I worked for close to 4 years at Facebook AI Research, where I helped push forward the state-of-the-art in neural machine translation and improved our understanding of how to train deep neural nets. I completed my Ph.D. at U. of Montreal with Yoshua Bengio on the subject of scaling deep learning algorithms. My collaborators and I have won two international AI competitions with our algorithms: the Unsupervised Transfer Learning Challenge in 2011, and the EmotiW challenge in 2014.


  • Research

    Here are selected publications. Full list available on Google Scholar.

    Published in Neural Information Processing Systems 2014

    This paper dispels the myth of bad local minima in high dimension and shows that the loss surfaces of neural networks have remarkable properties. This paper has helped renew interest for non-convex optimization as well as the study of neural nets using methods from statistical physics.

    Published in Neural Information Processing Systems 2011

    Accepted as oral presentation (0.1% of submissions)

    We propose a classifier that learns to be invariant to transformations that are discovered without supervision. We show that the method learns to be invariant to transformations like translation and rotation in the case of vision problems. At the time of publication, we set new state-of-the-art results with this method.

    Published in IEEE/ACM Transactions on Audio, Speech, and Language Processing 2015

    Semantic slot filling is one of the most challenging problems in spoken language understanding (SLU). In this paper, we propose to use recurrent neural networks (RNNs) for this task, and present several novel architectures designed to efficiently model past and future temporal dependencies achieving state-of-the-art results.

  • Press

    Here are selected articles in the press about my work.

  • Talks

    At the International Conference on Machine Learning (ICML) 2016.


    At the Neural Information Processing Systems (NIPS) 2011.