Dr. Dmitrii Todorov's research group

Group Research, seen through three different perspectives:

[PhD position I had until recently is already filled]

From the neuroscience perspective:

Motor control is important for all aspects of our life. In humans it is studied mostly either in purely medical setting or using purely behavioral measurements. Yet its mechanisms are known to have origins in the brain. Neural data of good enough quality in humans in relation to motor control is not yet available so one needs to supplement existing data with computational modeling / machine learning with right inductive biases. My groups works on studying motor control (and specifically motor adaptation) from various angles: 

• designing and conducting neuro-behavioral experiments (in particular, using MEG and EEG),

• developing and implementing advanced data analysis pipelines (including deep learning) for data from such experiments,

• developing mathematical foundations for such data analysis methods,

• also developing computational models for the phenomena observed in such data

From the data science perspective:

We develop methods for analysis (classification, regression, forecasting, etc) of spatially-organized multivariate time-series with low signal-to-noise ratio. We then spend significant time interpreting the fitted machine learning models in the context on neuroscience.

From the applied math perspective:

We look for ways to fit a parameterized high-dimensional partially observed stochastic process (e.g. a stochastic differential equation or a Hawkes process) to various features of data coming from applications in neuroscience. The data often has strong oscillatory properties (so the resulting process should have them too). We then spend significant time connecting the resulting process' properties with neuroscientific concepts.