Research Themes

Accurate models require understanding how the data is acquired and how the measured variables relate to the underlying physical system. Brain data is generally an indirect measure of the underlying neural mechanisms. We develop optimization frameworks to estimate the latent neural states and model parameters from high-dimensional data.

Complex dynamics drive brain activity across scales from EEG rhythms to cellular potentials. However, the computational relevance of these patterns remains elusive. We develop new approaches to elucidate the structure and significance of brain dynamics.

People differ, brains differ. Translating from theory to practice often hinges on this point. Our group emphasizes person-specific ("precision") brain modeling at every-level of analysis. We develop new algorithms to build detailed models from person-specific data. These models afford new insights into the science of individual differences and furnish the high accuracy needed for model-based engineering.

We aim to translate individualized brain models into treatments with particular emphasis on noninvasive brain stimulation. Thus, we seek optimal neurostimulation waveforms (inputs) or policies (feedback) to improve cognitive function using control-engineering. However, classical techniques are challenged by high-dimensional, nonlinear systems. Moreover, it remains unclear how to mathematically frame a control-objective that spans diverse cognitive processes. We are working towards geometric objective-framing to treat these problems as well as new control paradigms to reconcile cellular computation with spatially-coarse stimulation.

Cognitive control refers to the process of altering behavior in response to new contexts and internal goals. This capacity is fundamental to human intelligence and remains challenging to model or implement in artificial systems (e.g. neural networks). We study cognitive control across scales: how do spatially-fine neural representations alter brain-wide activity? How does high-frequency neuronal firing direct extended behavior?