Moustapha Cisse


I am a scientist at Facebook Artificial Intelligence Research (FAIR). I am committed to building an axiological artificial intelligence to improve our society, particularly the lives of those who need it most. My current research efforts focus on essential prerequisites of such AI: Fairness, Transparency, and Reliability. The Newscientist and the MIT Technology Review have highlighted our recent work in this area. I was born and raised in the beautiful Senegal (you should visit!), where I studied maths and physics at Universite Gaston Berger. Later, I completed a Ph.D. in machine learning at Universite Pierre et Marie Curie in Paris (France).


Selected Recent Publications

Mixup: Beyong Empirical Risk Minimization (preprint, 2017)

H.Zhang, M. Cisse, Y. Dauphin, D. Lopez-Paz··full paper

Convnets and Imagenet Beyond Accuracy: Adversarial Examples, Explanations, Model Criticism and Biases (preprint, 2017)

P. Stock, M. Cisse··full paper

Countering Adversarial Images using Input Transformations (preprint, 2017)

C. Guo, M. Rana, M. Cisse, L. Maaten··full paper

Kronecker Recurrent Units (preprint, 2017)

C. Jose, M. Cisse, F. Fleuret··full paper

Houdini: Fooling Deep Structured Visual and Speech Recognition Models (NIPS 2017)

M. Cisse, T. Adi, N. Neverova, J. Keshet··full paper

Unbounded Cache Model for Online Language Modeling With Open Vocabulary (NIPS 2017)

E. Grave, M. Cisse, A. Joulin··full paper

Parseval Networks: Improving Robustness to Adversarial Examples (ICML 2017)

M. Cisse, P. Bojanowski, E. Grave, Y. Dauphin, N. Usunier··full paper

Efficient softmax approximation for GPUs (ICML, 2017)

E. Grave, A. Joulin, M. Cissé, D. Grangier, H. Jégou··full paper