Pereira, Francisco Pereira. You can email me at [first].[last][at]gmail[dot]com .


I lead the Machine Learning team at the National Institute of Mental Health. This page is no longer being updated, please look there for more recent work.

Prior to this, I was a researcher at Medical Imaging Technologies, Siemens Healthcare, where I managed the Computational Neuroscience program. I was also the PI of a team in the IARPA Knowledge Representation in Neural Systems program. I was a postdoc in the Botvinick Lab (and a frequent lurker in the Computational Memory Lab) at the Princeton Neuroscience Institute. I got my Ph.D. in the Computer Science Department at CMU, working with Tom Mitchell and Geoff Gordon. I was also a student in the graduate training program of the Center for the Neural Basis of Cognition, working in collaboration with the Center for Cognitive Brain Imaging. I got my undergraduate degree at the Computer Science Department of the University of Porto, in Portugal.


what's in the box?
My main area of research is the study of how semantic knowledge is represented in the brain, using machine learning and brain imaging. I work on methods to extract semantic information from text corpora and combine it with structured knowledge databases, in order to build models of human performance on various semantic tasks. I also work on methods to relate models of semantic mental processes to brain imaging data, for validating those models through brain decoding tasks.

In parallel with this, I have an interest in developing new machine learning approaches for answering clinical questions — diagnostic and prognostic — using brain imaging data (functional, structural and diffusion MRI, as well as spectroscopy), and for automatically providing explanations and context for those answers in terms of models of the underlying condition.

If you want to know more about my work


(my Google Scholar page, should you want to cite any :)

open-source software

These MATLAB+C toolboxes are geared towards doing "searchlight"-style analyses over the entire brain, considering all overlapping small voxel neighbourhoods, in seconds: Other toolboxes:

invited presentations