Slavakis Lab

AboutAbout the lab

Exploiting the geometric structure hidden in data.

The Slavakis Lab studies signal processing and machine learning, focusing on the design of computational methods (numerical algorithms).
The lab aims to extract and exploit the geometric structure underlying signals and data, in order to build signal processing and machine learning frameworks that are more flexible than before.
Its applications span a wide range of fields, including reinforcement learning, medical imaging, brain networks, and quantum machine learning.
Because a background in mathematics is important for the research, the lab welcomes visits and inquiries from students who are considering joining the lab or going on to its graduate program.

ResearchResearch themes

The lab designs computational methods for signal processing and machine learning. It works on a wide range of themes and applications, including reinforcement learning, learning on manifolds, optimization using Riemannian geometry and monotone operators, quantum signal processing, medical imaging, and analysis of network time series.

*Illustrative image

New reinforcement learning algorithms↗

The lab designs new reinforcement learning algorithms that work when states and actions are continuous and high-dimensional. They incorporate mechanisms that require no information about transition probabilities and can run without training data.

*Illustrative image

Reducing bias in sparse optimization↗

The lab studies methods to reduce the estimation bias introduced by the widely used L1-norm approach in sparse optimization, which seeks solutions in which most components are zero. It has also proposed a new concept called the "external division operator."

*Illustrative image

Regression on manifolds↗

The lab is building a regression framework that learns from data lying on a low-dimensional manifold. Assuming only that the manifold is smooth, it applies the framework to problems of imputing missing data in dynamic MRI and graph signal processing.

*Illustrative image

Classifying data on Riemannian manifolds↗

The lab studies methods for clustering and classifying data on Riemannian manifolds, a special class of manifolds on which distances can be measured. It applies them to network time-series data such as brain networks.

VoicesStudent voices

Ratings and reviews reflect the experiences of individual current or former students, not the lab's official views. Whether a lab has reviews, or how many, says nothing about the lab itself.

Students who take part in an interview may receive a thank-you gift worth ¥1,000, regardless of what they say.