Systems level Bayesian models of the brain suggest that activity in cortical hierarchies conforms to a predictive coding architecture. Each level in the hierarchy is located in a different brain region and each region has a population of "error units" and a population of "causal units". Neocortex itself has a finite thickness, with some cells in deep laminae (closer to white matter) and some more superficial. The error units are thought to reside in superficial cortical laminae and causal units in deep laminae. Error units receive messages from the state units in the same level and the level above, whereas state units are driven by error units in the same level and the level below. The person near the centre of the image below (can you see them ?) would be difficult to see without a top-down prediction. This prediction may be derived from previous time steps, hence the need for dynamic models, or from higher-level scene knowledge that images of outdoor scenes contain paths and people walk on paths.
In collaborations with Emrah Duzel's group we have shown eg. that memory representations are reactivated during the delay period of a working memory task. More amazingly, the reactivations occur periodically, in line with the brains 'theta cycle'(4-8Hz). The more tightly the reactivations are synchronised to theta, the better a person's memory.
The ability of mammals to navigate is well studied, both behaviourally and in terms on the underlying neurophysiology. Navigation is a well studied topic in computational fields such as machine learning and signal processing. However, studies in computational neuroscience, which draw together these findings, have mainly focused on specific navigation tasks such as spatial localisation. In a series of papers, we propose a single probabilistic model which can support multiple tasks, from working out which environment you are in, to computing a state trajectory that will take you to the desired goal. We describe how these tasks can be implemented using a common set of lower level algorithms that implement "forward and backward inference over time" and propose that they are related to recent experimental findings of "pattern replay".
In collaborations with Sara Bengtsson's group we have studied eg. the effect of self esteem, operationalised using semantic priming, on performance of a rule-switching task. Normally, when people make a mistake on a task they slow down on the next trial. We found that when people were primed with associations to the word 'stupid' they slowed down after a correct response ! We were able to explain this phenomenon, using a standard evidence accumulation model but with an adaptive threshold set according to Bayes' rule.
It is possible to assess the tuning curve of populations of sensory neurons using behavioural measures or fMRI. The images below show amplitude and tuning centres/widths of populations of cells in the human auditory cortex as identified using fMRI (Kumar et al, 2014).