Razvan Pascanu
Research Scientist at DeepMind; contributor to Theano and first author of a study of RNN training difficulties
- Research Scientist, DeepMind (Google DeepMind)1
- Affiliate Member, Mila1
Razvan Pascanu is a Research Scientist at DeepMind and an Affiliate Member at Mila. Pascanu was involved in developing Theano, helped write some of its deep learning tutorials, and is a co-author of the 2012 paper on Theano's new features and speed improvements. Pascanu is first author of "On the difficulty of training Recurrent Neural Networks", which analyzed exploding and vanishing gradients and proposed norm-based gradient clipping. Pascanu also co-organizes EEML and has helped launch or organize conferences on lifelong learning and on learning on graphs.132
Work
On the difficulty of training Recurrent Neural Networks (external site: arxiv.org)Published research
First author, with Tomas Mikolov and Yoshua Bengio, of the 2012 paper proposing norm-based gradient clipping for exploding gradients and a soft constraint for vanishing gradients.2
Theano (external site: github.com)Open-source code
Involved in developing Theano and in writing some of its deep learning tutorials; co-author of "Theano: new features and speed improvements" (2012).134
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.