My Research
I find staggering the Human brain’s ability to make choices that are more flexible, efficient and robust than today’s most powered machines, in a bewilderingly complex world and under scarce resource. The overarching goal of my research is to understand the computational principles and limitations that underlie Human adaptive and efficient decision-making.
Focus: I study how neural networks integrate context with sensory information to inform decision-making, within and across brain regions, by combining high density recording & fMRI techniques, psychophysics, data-constrained computational modeling. My research focuses on the following questions:- How does the brain efficiently encode continuously changing external sensory inputs and internal context in the dynamic activity of cortical neural population?
- How are they abstracted into internal models that inform decision-making?
- Are they fused in cortical networks? If yes, via what computations?
- How does the cortical circuitry's sparse, structured connectivity, constrain its representational and computational capacity?
- What can we learn from such constrains on the limitations of animal perception and decision-making?
Approach: I combine system neuroscience neural recording techniques, such as extracellular electrophysiology in rodents and fMRI in humans, with psychophysical experiments, to jointly measure neural population activity and behavior and I use decision theoretic and neural network models to link them. To thoroughly describe the dynamic computations reflected in the recorded activity, I analyse how neural network connectivity pattern shapes and constrains the representational and computational capacity of recurrent neural network models of the cortical microcircuit at different levels of biophysical detail. To do so I use mathematical tools from linear dynamical system theory and machine learning techniques of dimensionality reduction. High-density neuropixels multi-unit recordings combined with machine learning-driven progress in spike sorting, have reduced sampling biases, such as the preferential selection of task-responsive neurons. Yet the amount and types of sorting biases that limit our ability to measure the true population activity are unclear. I use biophysically-detailed models of the neocortical microcircuit to characterize and quantify spike sorting biases and improve spike sorting algorithms.
Some context: I currently lead research as a postdoc in Michael Reimann's lab at the Blue Brain Project, EPFL. I create experimental, computational approaches and machine learning tools to analyse stimulus and context encoding in large-scale biophysically-detailed simulations of the entire primary somatosensory cortex of rodents, relying on the computing powers of high-performance clusters and graphical processing units (GPU). Before joining the Blue Brain Project, I worked with Prof. Justin Gardner at Stanford University, where I used psychophysics, fMRI, and Bayesian modeling to link human visual choices with cortical activity. I studied Computational Neuroscience at the University of Bordeaux, France, where I obtained my PhD under the supervision of Prof. Thomas Boraud. During my PhD, I performed multi-electrode array (MEA) recordings in the Basal Ganglia of primates trained to perform a stochastically rewarded two-choice task. I specialized in developing statistical approaches and reinforcement learning models to link the animal choices with the measured MEA spiking activity. I found that choices are not always motivated by reward but are sometimes entirely driven by task contextual cues. I pursued this research first at the Riken Brain Science Institute in Japan, then at Stanford University. During this time, I honed my skills in machine learning and statistical decision modeling and acquired new skills in fMRI to link human decision-making, prior experience, and cortical activity. I then spent a brief period in industry working as a Lead Data Scientist where I perfected my machine learning engineering skills. I currently also pursue research as a visiting scholar at KU Leuven, where I investigate how artificial networks can self-organize to represent information efficiently and make adaptive decisions in collaboration with Prof. Cees Van Leeuwen.
My CV
Academic appointments
- Swiss Federal Institute of Technology, S., Blue Brain Project | Connectomics lab. Postdoctoral scientist (2022 - now)
- KU Leuven, Brain and Cognition | Lab. of Cees Van Leuven (2021 - now)
- Stanford University, Dept. of Psychology | Lab of Justin L. Gardner, Postdoctoral scientist (2014 - 2017)
- Riken Brain Science Institute | Lab. of Hiroyuki Nakahara and Lab of Justin L. Gardner, Postdoctoral scientist (2010 - 2014)
- CNRS, Movement Adaptation Cognition | Lab. of Thomas Boraud, France Parkinson Doctoral Fellow (2007 - 2010)
- CNRS, Neurobiology of Behavior then Movement Adaptation Cognition units | Lab. of Yoon Cho and Lab. of Abdelhamid Benazzouz, Graduate Researcher (2006 - 2007)
Industry appointments
- Societe General for Capgemini Engineering, S.,Lead Data Scientist | Paris, France (2020 - 2021)
- Altran, Senior Data Scientist | Paris, France (2019 - 2020)
- Fundvisory , Lead research scientist | Paris, France (2018 - 2019)
Awards & Fellowships
- Stanford University Center for Cognitive and Neurobiological Imaging innovation grant (2014', 2015)
- Japan Society for the Promotion of Science fellowship (JSPS, I declined it, 2010)
- France Parkinson Doctoral Fellowship (2009 - 2010)
- College merit scholarship (2002 - 2007)
Peer-reviewed publications
- Laquitaine, S., Imbeni, M., Tharayil J., & Reimann M., (2023). A biophysically detailed cortical circuit model to evaluate and improve spike sorting of large-scale neural recordings (In preparation).
- Rentzeperis, I., Laquitaine, S., & van Leeuwen, C. (2022). Adaptive rewiring of random neural networks generates convergent–divergent units. Communications in Nonlinear Science and Numerical Simulation, 107, 106135
- Laquitaine, S., & Gardner, J. L. (2018). A switching observer for human perceptual estimation. Neuron, 97(2), 462-474.
- Laquitaine, S., Piron, C., Abellanas, D., Loewenstein, Y., & Boraud, T. (2013). Complex population response of dorsal putamen neurons predicts the ability to learn. PLoS One, 8(11), e80683.
- Chetrit, J., Ballion, B., Laquitaine, S., Belujon, P., Morin, S., Taupignon, A., ... & Benazzouz, A. (2009). Involvement of Basal Ganglia network in motor disabilities induced by typical antipsychotics. PLoS One, 4(7), e6208.
Teaching
- Cajal Advanced Training , Traditional and machine learning techniques to analyse high-dimensional neural and behavioral data (Instructor, upcoming spring, 2024)
- Neuromatch Academy 2023 , Computational Neurosciences, Behavior and Theory , created Computational Neuroscience project resource material (spring, 2023)
- Neuromatch Academy 2022 , Computational Neurosciences, Behavior and Theory , Lead Project TA (summer, 2022, 40 students)
- How to develop and productionize reproducible and maintainable machine learning models, Advanced training in Data Science Industrialization (10-20 practitioners, Societe Generale Bank (three quarters, 2020 & 2021), lead instructor and course designer
- Machine learning model monitoring, Advanced training (10 practitioners), Societe Generale bank (2 sessions, 2020-2021), lecturer
- Unsupervised intent recognition, Advanced training (10 practitioners), Hybrid Intelligence, Capgemini Engineering (2021), led research & designed lecture
- Humans approximate Bayesian inference with a switching heuristic during motion direction estimation, Stanford Psychology department, Memory-Decision seminar series (10 participants, 2016), lecturer
- Functional connectivity of visual cortex in the blind follows retinotopic organization principles, Stanford Psychology Department, Vision seminar series (10 participants, 2015), lecturer
- The Scientific Methodology, Introductory Undergraduate (10-20 students), Bordeaux University, Winter quarter (2007-2009), Teaching assistant
Seminars
- Stanford University Department of Psychology, Human approximate Bayesian inference with switching heuristic during motion direction estimation. Seminar, Palo Alto, CA, Nov 5, 2014
- RIKEN BSI Ko, H. et al. The emergence of functional microcircuits in visual cortex. Nature 496, 96–100 (2013). Seminar, Wakoshi Japan, April 2, 2014
- RIKEN BSI Ko, H. et al. RIKEN BSI Adesnik, H., Bruns, W., Taniguchi, H., Huang, Z. J. & Scanziani, M. A neural circuit for spatial summation in visual cortex. Nature 490, 226–231 (2012). Seminar, Wakoshi Japan, Oct 2, 2013
- RIKEN Brain Science Institute (BSI) Humans exploit uncertainty and bimodality of priors in motion direction estimation, Conference talk, Wakoshi Japan, Sept 19, 2013
- Meeting of Bordeaux Laboratories Dorsolateral Putamen neurons predict irrational preferences in a motivated choice learning. Conference talk, Bordeaux France, April 15, 2010
- Conjoint meeting CGB - CMA Action selection reinforcement learning in the Basal Ganglia: an electrophysiological and computational approach in the non human primate. Conference talk, Bordeaux France, May 25, 2009
- France-Israel Neuro-Robotics Meeting, Investigating the synaptic and neuronal mechanisms in the BG that contribute to operant conditioning. Conference talk, Jerusalem Israel, Nov 25, 2008
- Conjoint meeting CGB - CMA Modulation of SNr activity with a DRD5 inverse agonist: perspective to the physiopathology of Parkinson Disease. Conference talk, Croisic France, Jun 8, 2007
Peer-reviewed International Conference Proceedings
- Rentzeperis I., Laquitaine, S., Cees van Leeuwen (2020, January). Adaptive rewiring evolves brain-like structure in directed networks. In CEUR Workshop. Madrid, Spain.
- Laquitaine, S., & Gardner, J. L., (2013) Humans exploit the uncertainty in priors to improve direction perception. In Cosyne, Salt Lake City, Utah, February 8th - March 3
Peer-reviewed national Conference Proceedings
- Laquitaine S., Loewenstein Y., Gross C., Hansel D., Boraud. (2009) T. Learning behaviour in a two alternative decision task in primate follows a gradient-based learning rule. Neurocomp’09, Computational Neuroscience : From Multiple levels to Multi-level, page 19, Bordeaux, France
Non peer-reviewed
- Birman D., Cable, D., & Laquitaine S. (2017) Applying 3D Convolutional Neural Networks to Human Psychophysics, Working Paper for Stanford CS231N, Palo Alto, CA, USA
- Rentzeperis I., Laquitaine S., Gardner J., (2016) Expectation modulates activity in Visual Cortex, Conference abstract at AREADNE, Santorini, Greece.
- Laquitaine S., & Justin L. Gardner. Humans exploit the uncertainty and the bimodality of their priors to improve direction perception (2013), at BSI retreat, Karuizawa, 2013; at SFN, San Diego, November;at Neuro2013, Kyoto, Japan, June 22th. At Cosyne 2013, Salt Lake city, Utah, February 8th - March 3).
- Kaveri S, Koene A, Laquitaine S., Nakahara H (2012) Sign-tracking and goal-tracking by a reinforcement learning model. Conference abstract at The 42nd Annual Meeting of the Society for Neuroscience, New Orleans, LA, USA, October 13-17, 2012
- Laquitaine S., Loewenstein Y., Gross C., Hansel D., Boraud T. Learning behaviour in a two alternative decision task in primate follows a linear single free parameter probability matching law. 9th colloquium of Société des Neurosciences, Bordeaux, France, 2009.
- Laquitaine S., Loewenstein Y., Boraud T. Learning of the two-armed bandit reward schedule by primates reveals complex choice strategies underlying different neural activity patterns. Society for Neuroscience, October, 2009
- Chetrit J., Ballion B., Laquitaine S., Belujon P., Morin S., Gonon F., Taupignon A., Bioulac B., Gross C.E. and Benazzouz A. Hypokinesia and catalepsy induced by flupentixol are correlated with changes in neuronal activities of the basal ganglia. Society for Neuroscience, Washington, Nov. 15-19, 2008.
- Delcasso S., De Rivoyre J., Laquitaine S., Jeantet Y., Cho H.Y. The use of an operant DRL schedule for longitudinal studies of APP transgenic mice. Society For Neuroscience, Atlanta, 2006.
Open source software
Self-organizing networks
- Simulation code for "Rentzeperis, Laquitaine & van Leeuwen, CNSNS (2022)": simulations & analysis of neural network rewiring into an efficient recurrent convergent-divergent architecture
Visual inference
- bsfit: a software with models, heuristics & analytic tools to study circular estimate data under the framework of Bayesian statistical decision theory
- code for "Laquitaine & Gardner, Neuron (2018)": Bayesian and heuristic model of perceptual estimates
fMRI voxel modeling
- projBrainInference: computational modeling & analysis of the cortical correlates of motion direction inference
- voxppdec: probabilistic population decoding of fMRI voxel signals
- slforwmodeling forward channel modeling of fMRI voxel signals
Deep learning for image processing
- motnet: a 3D Convolutional Neural Networks for Human Psychophysics
- convnet: a deployable load-balanced convnet software for image classification with tensorflow, docker and traefik
Deep learning for language modeling
- vad: a Long short-term memory recurrent neural network model of voice activity detection
- intent: a proof of concept model for unsupervised intent recognition
- slSR: the productionization of slotRefine, a state-of-the-art multi-headed attention model for intent classification and slot filling
My contributions to large collaborations:
- mrTools: (Home, Zenodo): A package for the analysis and visualization of functional magnetic resonance imaging data
Mentoring & acknowledgements
- A Quantitative Framework For Motion Visibility In Human Cortex, Birman D & Gardner JL, Journal of Neurophysiology (2018)
- Adaptable History Biases In Human Perceptual Decisions, PNAS, Abrahamyan A, Silva LL, Dakin SC, Carandini M, Gardner JL (2016)
- Distributed Processing Of Color And Form In The Visual Cortex, Rentzeperis I, Nikolaev AR., Kiper DC., and Cees van Leeuwin, Front Psychol (2014)
- Dan Birman's Thesis, Computational Linking Models Of Human Selective Visual Attention, Stanford University, 2019
- Cameron Mackenzie's Thesis, The Neural Basis of Tactal Attention, Stanford University, 2018
- Siva Kaveri’s Thesis, Japan, 2015 (http://bit.ly/1kOiene);
- Aude Retailleau’s thesis, Bordeaux, 2015, (http://bit.ly/1SRfFLD);
- Rachida Ammari's Thesis, Bordeaux, 2010 (http://bit.ly/1HfiAxh).
- Camille Piron's Thesis, Bordeaux, 2014 (http://bit.ly/2BV7c81)