Petar Djuric
Stony Brook University, NY·Engineering
Many systems in medicine and engineering can only be understood through the signals they produce over time or the measurements taken inside them. One example is the electrical activity recorded from a brain with epilepsy.
- Fields
Engineering
Computer Science
Mathematics & Statistics
- Topics
signal processing
machine learning
Bayesian inference
- Grants
- 2 known grants, through Jun 2027
What they work on
- Develop deep state-space models that track hidden system dynamics from multivariate time signals, using Bayesian estimation that needs little data and few assumptions about the system.
- Extend these models to regime switching, in which a system jumps between distinct modes, and develop methods to estimate model structure from data.
- Develop theory and methods for causal discovery among multivariate time signals to determine which signals drive which.
- Identify the brain states and zones that cause seizures by applying the new models to local field potentials, electrical signals recorded inside the brains of epilepsy patients.
- Design passive radio frequency sensors embedded in concrete that communicate by backscatter and are powered by solar-powered exciters, so they can sense deformation, temperature, humidity and damage.
- Use ensembles of Gaussian processes, flexible statistical models, to let embedded sensors locate themselves and to map 3D distributions of material properties inside large concrete structures.
More about the research
Another is the temperature, moisture and strain inside a concrete bridge. Turning such data into reliable knowledge requires good models, and also methods that report how uncertain their conclusions are. Across two projects, the lab develops probabilistic machine-learning and signal-processing methods for these problems.
Why it matters
For people with epilepsy, surgically removing the brain zones that cause seizures often cures them. Methods that locate those zones more reliably from brain recordings could help neurosurgeons plan treatment. For aging bridges, buildings, pipelines and other concrete infrastructure, embedded sensor networks could give long-term, high-resolution warning of damage. This could help keep these structures safe as urbanization and climate change add strain.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Aatmesh Shrivastava
Northeastern University, MA·Engineering·Early-career
Develop a distributed beamforming technique that combines RF energy from multiple Wi-Fi or Bluetooth-type transmitters and directs it to a remote IoT receiver, using closed-loop optimization between transmitters and receiver.
Kaijian Liu
Stevens Institute of Technology, NJ·Engineering·Early-career
Develop a learning framework that estimates the conditions of people and infrastructure, across space and time, from sparse and uncertain observations, as a basis for modeling disaster disruptions.
Georgios Giannakis
University of Minnesota, MN·Engineering
Model the resonant-beam optical wireless channel, an infrared beam formed in an open-air optical cavity, using nonlinear time-varying Volterra series and kernel-based learning for both downlink and uplink.
The school
Stony Brook University: 32 labs on PI Pages, $319M a year on research, 289 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.