Sensory Information Processing Lab

what we do

Information is one of the most valuable commodities in the world, and systems often face the task of trying to manage large data volumes while uncovering the most relevant information. This is true in the case of biological neural systems trying to use their senses to understand the environment. It is also true for engineered systems where we observe and manipulate complex dynamic systems.

A primary goal of the SIPLab is to develop algorithmic foundations for interfacing (directly or indirectly) with neural systems at scales from single cells to human perception. This work spans foundational contributions in machine learning, signal processing, computational vision and neuroengineering. Pursuing these goals results in novel algorithms and analysis for large-scale data science that can be used in machine learning applications beyond neuroscience. Application areas are always evolving, but currently include:

  • computational and theoretical neuroengineering in several systems (e.g., vision, audition, somatosensation) and paradigms (e.g., electrophysiology, artificial stimulation, neural prosthetics, closed-loop control);
  • computer vision and machine learning, especially involving human-computer interaction;
  • dimensionality reduction, manifold learning and dynamic filtering;
  • imaging and remote sensing (including hyperspectral imagery);
  • brain-computer interfaces; and
  • novel neuromimetic computing approaches.

We approach these various applications using the same core intellectual ideas. First, even when data is high-dimensional, the information of interest is generally low-dimensional and can be represented using geometric structures (e.g., linear subspaces, sparsity models, manifolds, and dynamical system attractors). Second, interesting systems generally have some type of feedback or closed loop behavior, especially when the systems involve interactions between machines and biology.

Please see our publications for more information on our current research outcomes. SIPLab members are affiliated with the Neural Engineering Center and the Center for Signal and Information Processing at Georgia Tech.


Cosyne workshop

Join us at the Cosyne 2018 workshop we're organizing (with Garrett Stanley) on "Closed-loop control of neural systems and circuits for scientific discovery". Exciting topic and great speaker lineup. Student travel grants available thanks to IEEE Brain. Click "read more" for details.

Hiring postdoc

SIPLab is now hiring a postdoc position in time-series data analysis (machine learning and signal processing) for a clinical project on deep brain stimulation for depression. Click "read more" to see the full ad.

Paper published in IEEE Trans. Image Processing

The paper "Cell Membrane Tracking in Living Brain Tissue Using Differential Interference Contrast Microscopy" by Lee, Kolb, Forest and Rozell shows a novel algorithm that can track neuron membranes in brain slices with very high accuracy in real time, representing a critical step on the way to automated high-throughput electrophysiology. Congrats John!

Four SfN presentations

SIPLab members will be presenting four posters at the upcoming Society for Neuroscience Annual Meeting. Stop by each day Sunday through Wednesday to learn about our recent results in neuro theory, neurotechnology and brain-machine interfaces!

Manivasagam Wins GT research award

Siva Manivasagam was recently awarded the GT University Interdisciplinary Research Award for a proposed collaboration with Chethan Pandarinath's lab. Congratulations Siva!

Paper accepted at International Conference On Machine Learning And Applications (ICMLA)

The paper "Rank Learning by Ordinal Gerrymandering" by Fenu and Rozell provides a new boosting approach to dimensionality reduction when the goal is to preserve the rank order of the data. Congrats Stefano!