2 Open PhD positions to investigate the neural basis of learning and working memory.
We are currently looking for two highly motivated graduate students to carry out their PhD research in our laboratory. Our lab offers an interdisciplinary stimulating environment, that values rigor, collaboration, close mentoring and critical but positive thinking. We are also a active member of the Barcelona Cognitive Computational and Systems Neuroscience community (barccsyn.org) which connects a large number of groups in Barcelona with common interests and complementary approaches.
The projects proposed are the following:
Project “Circuit mechanisms of priors and learning during decision making“ . This PhD work will be part of the NIH-AEI Collaborative Research in Comptuational Neuroscience (CRCNS) program and will be carried out in our lab in collaboration with Robert G. Yang (MIT) and Manuel Molano (CRM).
The candidate should have a masters in Neuroscience (or similar), experience working with mice or rats, fluent English, ability to work in an interdisciplinary team, be responsible and oriented to the detail. We will value programming skills in Python and R, experience doing stereotactic surgeries and writing/communication skills.
The project consists in studying how rats and humans learn simple categorization tasks with different statistical structure (i.e. different serial correlations in the stimulus sequence). We will be performing population recordings in behaving animals, optogenetic experiments and manipulations of synaptic plasticity during task learning.We offer a full time three-year position with a competitive PhD salary. Starting date will be between February and June 2023. Candidates interested in the position, please send an email to Jaime de la Rocha.
Project “The neural basis of working memory history biases as components of statistical learning” funded by the AEI.
The candidate should have a Master’s Degree in Neuroscience, Cognitive Science, Data Science or similar, have good programming skills, be fluent in English and have a genuine interest for cognitive research from a quantitative biological approach.
The project investigates the neural basis of history biases in working memory and their relation to sequence prediction using combined computational modeling, analysis of neurophysiological data from monkey experiments, and EEG and intracranial recording experiments in humans. See project summary below.
We offer a full time four-year position with a competitive PhD salary. Starting date is negotiable, but should be before September 2023.
Candidates interested in the position, please send an email to Albert Compte.
Summary “Circuit mechanisms of priors and learning during decision making“ :
The use of perceptual decision making tasks in animals has propelled our understanding of the neural basis of elemental cognition. However, we still lack a fundamental understanding of the mechanisms at play during task learning. Animal training remains a venturesome enterprise commonly resulting in suboptimal behaviors plagued with superstitious ticks and idiosyncratic biases. One prominent example of such suboptimality are sequential effects: animals tend to bias their choices based on previous decisions and outcomes, hindering performance in common laboratory tasks using independent trials.
Training recurrent neural networks (RNN) on decision making tasks has become a common approach to study potential neural mechanisms of cognition. Yet, RNNs typically behave much closer to optimality in laboratory tasks than real subjects. We suggest this behavioral difference is rooted in the fundamental discrepancy between how animals and current RNNs learn: unlike animals before learning, RNNs before training are tabula rasa and their connectivity is adjusted exclusively to the local contingencies of the task.
We hypothesize that animals’ learning of simple laboratory tasks builds mostly on pre-existing programs, namely structural prior, that have been shaped through evolution for the species’ fitness in a given ecological niche. Sequential effects are a manifestation of such pre-wired strategies, which may ultimately support learning, by the rapid formation of new associations between past events and future rewards. To test this, we will characterize sequential effects during learning of a set of perceptual tasks and identify their underlying neural circuitry. In close interaction, we will compare animals’ behavior with RNNs which, after being equipped with structural priors, can then mimic the animal’s ability to learn new tasks.
Summary “The neural basis of working memory history biases as components of statistical learning”:
Our brain excels at finding structure and meaning in incoming streams of sensory information without any explicit teaching signal. Such statistical learning underlies language and other high-order functions and has been shown to be defective in various mental disorders (e.g. autism, schizophrenia), but little progress has been made in understanding its neural network basis. This project aims to provide a neural network understanding of statistical learning by defining recently described attractive history biases in working memory as a proxy of statistical learning. By harnessing the strong theoretical basis of working memory, we will combine experiments in humans, data analysis of neural spiking data and computational modeling to test the hypotheses that short-term working memory history biases are supported by synaptic plasticity mechanisms operating in the local prefrontal circuit, while longer-term working memory history biases depend on cortico-cortical interactions with other brain areas, and these mechanisms contribute to statistical learning. To test this, we will pursue two aims: (1) define neural network dynamics supporting working memory history biases using computational models constrained by human electrophysiological and magnetic resonance imaging recordings; and (2) test the association of working memory history biases with sequence prediction in healthy statistical learning. By integrating data from brain structure, neural function and behaviour in humans and monkeys, we will deliver mechanistically consistent biophysical network models of statistical learning. This knowledge will foster understanding of the biology of sequential cognitive processing, and provide tools to advance towards mechanistically informative behavioural biomarkers that allow stratification of persons with deficits in statistical learning.