New perspectives and applications for greedy algorithms in machine learning

dc.contributor.advisorGhosh, Joydeep
dc.contributor.committeeMemberDimakis, Alexandros G
dc.contributor.committeeMemberShakkottai, Sanjay
dc.contributor.committeeMemberCaramanis, Constantine
dc.contributor.committeeMemberKoyejo, Oluwasanmi
dc.creatorKhanna, Rajiv Ashu
dc.date.accessioned2018-10-25T19:06:12Z
dc.date.available2018-10-25T19:06:12Z
dc.date.created2018-08
dc.date.issued2018-08-15
dc.date.submittedAugust 2018
dc.date.updated2018-10-25T19:06:12Z
dc.description.abstractApproximating probability densities is a core problem in Bayesian statistics, where the inference involves the computation of a posterior distribution. Variational Inference (VI) is a technique to approximate posterior distributions through optimization. It involves specifying a set of tractable densities, out of which the final approximation is to be chosen. While VI is traditionally motivated with the goal of tractability, the focus in this dissertation is to use Bayesian approximation to obtain parsimonious distributions. With this goal in mind, we develop greedy algorithm variants and study their theoretical properties by establishing novel connections of the resulting optimization problems in parsimonious VI with traditional studies in the discrete optimization literature. Specific realizations lead to efficient solutions for many sparse probabilistic models like Sparse regression, Sparse PCA, Sparse Collective Matrix Factorization (CMF) etc. For cases where existing results are insufficient to provide acceptable approximation guarantees, we extend the optimization results for some large scale algorithms to a much larger class of functions.The developed methods are applied to both simulated and real world datasets, including high dimensional functional Magnetic Resonance Imaging (fMRI) datasets, and to the real world tasks of interpreting data exploration and model predictions.
dc.description.departmentElectrical and Computer Engineering
dc.format.mimetypeapplication/pdf
dc.identifierdoi:10.15781/T24F1N387
dc.identifier.urihttp://hdl.handle.net/2152/69183
dc.subjectApproximate inference
dc.subjectSubmodularity
dc.titleNew perspectives and applications for greedy algorithms in machine learning
dc.typeThesis
dc.type.materialtext
thesis.degree.departmentElectrical and Computer Engineering
thesis.degree.disciplineElectrical and Computer Engineering
thesis.degree.grantorThe University of Texas at Austin
thesis.degree.levelDoctoral
thesis.degree.nameDoctor of Philosophy

Access full-text files

Original bundle

Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
KHANNA-DISSERTATION-2018.pdf
Size:
13.14 MB
Format:
Adobe Portable Document Format

License bundle

Now showing 1 - 2 of 2
No Thumbnail Available
Name:
PROQUEST_LICENSE.txt
Size:
4.45 KB
Format:
Plain Text
Description:
No Thumbnail Available
Name:
LICENSE.txt
Size:
1.84 KB
Format:
Plain Text
Description: