Xavier Boix
I build data-efficient algorithms and develop empirically-grounded theories of learning by bringing insights from biology to machine learning.
Research Scientist @ MIT
Xavier Boix
I develop biologically inspired Artificial Intelligence
An important feature of a learning machine is that its teacher will often be very largely ignorant of quite what is going on inside.
A. M. Turing, Computing Machinery and Intelligence (1950)
Learning is at the heart of both biological and artificial intelligence. Deep Neural Networks (DNNs) are one of the most successful biologically-inspired computational models. Yet, a similar success is still a distant goal in the domain of efficient learning mechanisms; there is a big gap between artificial and biological learning. The capability to learn with fewer training examples and overcome dataset bias is key for any intelligence to efficiently adapt to new environmental demands and to solve new complex tasks.

One of the main obstacles in developing biologically plausible learning models is the difficulty of understanding the internal functioning of learning machines, ie. the "black-box problem". I propose to use the scientific method on DNNs, ie. articulating hypothesis and test them with simulations as if they were another brain, to facilitate developing an understanding on DNNs.

My research lives at the intersection of engineering of DNNs, theoretical machine learning and neuroscience. My goal is to build biologically inspired DNNs while developing empirically-grounded theories that explain these models.
Currently, I work as a research scientist at MIT.

I feel fortunate to have received training in machine learning and neuroscience as a postdoc at MIT in the Sinha's lab and Poggio's lab, as well as the multidisciplinary NSF Center for Brains, Minds and Machines. Also, I obtained a doctorate from ETH Zurich (2014) in computer vision and completed a postdoc at the National University of Singapore (2015).
Publications and Code
Check out the full list of publications and code here:
Recent Projects
Here are some of my recent projects organized by theory (understanding DNNs with the scientific method), modeling (mimicking the brain) and applications (AI for social good).
1. Theory: Understanding DNNs with the Scientific Method
2. Modeling: Mimicking the Brain
3. Applications: AI for Social Good
The Language of Fake News

We open the "black-box" of DNN based fake news detectors. Our results show that the emergent DNNs' representations capture subtle but consistent differences in the language of fake and real news: signatures of exaggeration and other forms of rhetoric.
Past Projects
Here are some of my past projects completed during my PhD.
MIT 46-4077, 43 Vassar Street, MA 02139, Cambridge, USA


Made on