DUPOUX Emmanuel

Human inspired machine learning

emmanuel.dupoux [at] gmail.com

Emmanuel Dupoux

Short bio

Emmanuel Dupoux is full professor at the Ecole des Hautes Etudes en Sciences Sociales (EHESS), directs the Cognitive Machine Learning team at the Ecole Normale Supérieure (ENS) in Paris and INRIA (www.syntheticlearner.com) and is currently a part time scientist at Facebook AI Research. His education includes a PhD in Cognitive Science (EHESS), a MA in Computer Science (Orsay University) and a BA in Applied Mathematics (Pierre & Marie Curie University, ENS). He is the recipient of an Advanced ERC grant, the organizer of the Zero Resource Speech Challenge (2015, 2017, 2019) and the Intuitive Physics Benchmark (2019).

Topics of interest

Speech perception, language and cognitive development in infant, low resource language technology, automatic speech recognition, unsupervised and self supervised learning

Project in Prairie

Emmanuel Dupoux aims at reverse engineering how young children between 1 and 4 years of age learn from their environment, and construct machine learning algorithms that are more data efficient and flexible than current ones. He will develop unsupervised representation learning algorithms from raw audio or video, and evaluates them with cognitive developmental tests. He will study the inductive biases of neural architectures for language by studying how neural agents can develop communicative protocols. He will use these algorithms applied to naturalistic data to conduct quantitative studies of how infants learn across diverse environments.

Quote

Reverse engineering the ability of young children to learn languages is key to constructing machine learning algorithms that are more data efficient and flexible than current ones. It is also key to understanding how infants learn as a function of their input and to constructing predictive models for early diagnosis of developmental disorders.