FISH: Constraints and optimization for weakly supervised deep-learning from few data and fewer labels

Oct 23, 2020 - 12:00 PM to 1:00 PM EDT

Speaker: 

Prof. Bas Peters (Emory U.)

MIT Earth Resources Laboratory presents Bas Peters, Visiting Assistant Professor at Emory U., on "Constraints and optimization for weakly supervised deep-learning from few data and fewer labels."

"When pixel-level masks or partial annotations are not available for training neural networks for semantic segmentation, it is possible to use higher-level information in the form of bounding boxes, or image tags. In the imaging sciences, many applications do not have an object-background structure and bounding boxes are not available. Any available annotation typically comes from ground truth or domain experts. A direct way to train without masks is using prior knowledge on the size of objects/classes in the segmentation. I present a new algorithm to include such information via constraints on the network output, implemented via projection-based point-to-set distance functions. This type of distance functions always has the same functional form of the derivative, and avoids the need to adapt penalty functions to different constraints, as well as issues related to constraining properties typically associated with non-differentiable functions. I illustrate the capabilities in case of a) one or more classes do not have any annotation; b) there is no annotation at all; c) there are bounding boxes. Examples use data for hyperspectral time-lapse imaging, object segmentation in corrupted images, seismic interpretation, and sub-surface aquifer mapping from airborne-geophysical remote-sensing data. The examples verify that the developed methodology alleviates difficulties with annotating non-visual imagery for a range of experimental settings."

Bas Peters is a visiting assistant professor in the mathematics department at Emory University. Previously, Bas worked for Computational Geosciences Inc as a research scientist, and received his PhD degree from the University of British Columbia in 2019. His main research interests are constrained optimization; design, optimization, and regularization of deep neural networks, geoscientific and geospatial applications, inverse problems, image processing, and numerical linear algebra.

ERL's Friday Informal Seminar Hour (FISH) takes place most Fridays during the academic year, 12pm-1pm Eastern.