Machine learning, robotics, planning and learning in partially observable domains, reinforcement learning.

Image interpretation using planning and learning techniques; methods of theory formation for classification, configuration, planning and design problems with biomedical applications.

I develop computational models to learn generalizable and human-readable representations from unlabeled data, with a focus on natural language processing. To this end, I rely on mathematical frameworks such as Information theory: a representation is good if it transmits most information. Linear algebra: a representation is good if it lies in an optimal subspace. I am also interested in applications of learned representations to practical problems such as entity linking. I am not a theoretician by trade, but I enjoy working with theoreticians to study topics relevant to representation learning such as estimating mutual information.

Our lab is interested in the molecular mechanisms regulating dendritic spine morphogenesis and synaptic plasticity. Dendritic spines are small, actin-rich protrusions that receive most of the excitatory synaptic inputs in the brain.

Molecular mechanisms underlying sensory and cognitive function in mouse models of human diseases. The lab is interested in understanding how mutations in chromatin remodeling proteins results in hyperactivity and circling behavior in mutant mice. Identifying transcriptome changes in affected neurons by deep sequencing will help us understand how the activity and development of these neurons have been altered. Our goal is to bridge the molecular changes in neurons to the abnormal behavior observed in these mouse models.

Our lab models human neurodevelopmental disorders in mouse with a particular focus on Tourette Syndrome (TS). Despite the prevalence of TS in the general population (~1/100 individuals), the underlying pathophysiology is poorly understood. We are currently using CRISPR-based approaches to generate mouse models that harbor recently discovered human point mutations found in sporadic forms of TS. With these disease mouse models, we are investigating how circuit development and function in the brain are perturbed by the human mutations using a combination of mouse genetics, circuit labeling techniques, optogenetics, electrophysiology, and mouse behavior. Additionally, the mouse models will provide powerful tools for drug screening approaches that can inform complementary studies with human iPSC lines.

Philosophy and psychopathology (especially disorders of self); phenomenological philosophy; Wittgenstein; philosophical aspects of psychoanalysis.