PhDthesis

My thesis can be found here.

To know the motto of my research in a crux: Please have a look at this video link

    My PhD research is on computational modelling of the joint roles of neuromodulators dopamine and serotonin in decision making. We find that utility based decision making mediated by the basal ganglia with dopamine and serotonin controlling the reward and risk components, best explain their task specific manifestations in a variety of behavioral paradigms. My thesis can be found here. My PhD research was guided by Prof. V. Srinivasa Chakravarthy and Dr. Ravindran Balaraman

 Basal Ganglia is classically known for learning by reinforcement with the reinforcing signal believed to be dopamine. We, in our lab propose a Go-Explore-NoGo way of Basal Ganglia policy execution. For more on it: click here

 The classic way of understanding saliency in a reinforcement learning framework is through a surrogate of reward function called 'Value'. We realised the need to take up the notion of uncertainty to better determine the saliency in reinforcement framework. In the process, we noticed that the risk sensitive parameter in a risk based decision making framework also has the power of unifying the various views of serotonin under a single roof.

 Serotonin activity in Basal Ganglia is termed as 'complex' due to its diverse roles. The three primary roles of serotonin in Basal Ganglia, according to current thoughts on the subject, are: controlling the risk sensitivity, time scale of reward prediction and the punishment sensitivity of decision making. We constructed a model of the joint functioning of both serotonin and dopamine in Basal Ganglia, using concepts from Reinforcement Learning. In line with actor-critic approach to basal ganglia, it is assumed that dopamine activity represents the temporal difference error; the striatum to compute mean rewards or value;and  serotonin activity to be related to risk assessment (variance of the rewards). We then introduced a modified form of utility function that is a weighted summation of value and risk functions. Using this integrated formulation of the utility function, we showed the way the three diverse roles (risk sensitivity, time scale of reward prediction and punishment sensitivity) of serotonin can be reconciled under a single framework. For more on it: click here and here

Such task-specific manifestation of serotonin is found to be quintessential to explain the ON medicated PD patients excelling in reward learning. But these patients under perform in punishment learning. A vice-versa is observed in OFF-PD patients. 

 This concept of risk based decision making is further used to understand the increased precision grip exerted by Parkinson's Disease patients. To read more on it: click here

 Attempts to model the role of Acetylcholine in addition to existing functioning of dopamine and serotonin in Basal Ganglia, and using this detailed model to understand the gait of Parkinson's Disease patients (whose prime target is Basal Ganglia) is also carried on. To read more on it: click here