I am interested in exploring the computational principles and biological mechanisms underlying higher level cognition and intelligence in the human brain. How do we understand everything? How do we make sense of the information which we percieve on a daily basis? When we look at an object, how are we able to deduce all its attributes like colour, shape etc.? One possible hypothesis is that we are able to do these things due to the knowledge bank which we have acquired from our prior experiences. For example, we are able to infer the sense of a word having different meanings based on the context it is being used in. This is because the context provides additional information which helps us pick the correct meaning of the word from our knowledge bank. However, what if the given word doesn't exist in our knowledge bank (i.e., we have never seen that word before)? In this case, humans are still able to guess what the meaning of the word might be based on the contextual information provided. This indicates that human brain is not only able to recall what it had learnt before, but also use that information to make inferences.
Bayesian methods have been used to make computational models to explain such intelligence in humans, however most of the Bayesian models lack neural plausibility i.e., it is unclear how these computations might be implemented in the brain. This question of the neural basis of Bayesian computation is what I would specifically address in my research. I plan to build a biologically plausible computational model of Bayesian inference by mapping an existing Bayesian model to the Neural Engineering Framework (NEF). Besides inspecting the neural plausibility of a specific Bayesian model, I would also like to deduce a general approach for mapping any Bayesian model to the NEF.