Bahman Yari Saeed Khanloo


Contact: bahman[dot]yari[att-sign]gmail.com
Bahman Khanloo NICTA Monash University


About Me

My primary interest lies in learning theory where convex analysis, probability theory and information theory help serve machine learning community. I am also passionate about applying machine learning to non-trivial real-world problems where careful combination of theory and practice is essential.

Currently, I am with Monash University and National ICT Australia (NICTA) where I work with Dr. Reza Haffari and Prof. Wray Buntine.

I served as a teaching assistant in Leiden University and research employee at Center for Mathematics and Computer Science (CWI) in the Netherlands where I worked with Prof. Peter Grunwald as part of information-theoretic learning group.

Earlier, I did my Master's in computing science at Simon Fraser University (SFU) in Canada where I worked on convex optimization, graphical models and applications to computer vision under supervision of Dr. Greg Mori and Prof. Ghassan Hamarneh.


Selected Publications

A Bennett Inequality for the Missing Mass. (submitted)
[draft]
Novel Bernstein-like Deviation Bounds for the Missing Mass. 31st Conference on Uncertainty in Artificial Intelligence (UAI), 2015.
[pdf]
Novel Deviation Bounds for Mixture of Independent Bernoulli Variables with Application to the Missing Mass. (submitted)
[draft]

Summary: This is an instance of variance control problem where we shrink the variance while controlling the magnitude of terms. We improve state-of-the-art for both upper and lower deviation bounds on missing mass in the case of small deviations. We introduce a thresholding mechanism that helps reveal high concentration of missing mass around its mean for small deviation sizes by harnessing worst-case distributions that may otherwise inflate the variance. Interestingly, Lambert function plays a crucial role in our thresholding procedure.
A Large Margin Framework for Single Camera Offline Tracking with Hybrid Cues. Computer Vision and Image Understanding (CVIU), 2012.
(with Ferdinand Stefanus, Mani Ranjbar, Ze-Nian Li, Nicolas Saunier, Tarek Sayed and Greg Mori) [pdf] [more details]

Summary: We pose tracking as a structured input-output prediction problem. We devise a large-margin criterion for learning to combine static and dynamic information for tracking via trajectory optimization and build a system for automatic detection and tracking of pedestrians.


Research, Teaching and Training Experience


Relevant Courses


Useful Books and Resources


Software for Optimization and Machine Learning


Good Stuff


Favorite Quotes

"Everything was impossible until someone did it.", Unknown
"If you are not falling, you are not trying.", Sonnie Trotter
"You don't have to believe in God, but you should believe in The Book.", Paul Erdos
"Success consists of going from failure to failure without loss of enthusiasm.", Winston Churchill
"A mistake proves that someone stopped talking long enough to do something.", Phoenix Flame
"I would rather die of passion than of boredom.", Vincent van Gogh
"The best people possess a feeling for beauty, the courage to take risks, the discipline to tell the truth, the capacity for sacrifice. Ironically, their virtues make them vulnerable; they are often wounded, sometimes destroyed.", Ernest Hemingway
"There was nowhere to go but everywhere, so just keep on rolling under the stars.", Jack Kerouac

Copyright note: The articles in this page are presented to ensure timely dissemination of scholarly work and may not be (re)posted or published in whole or in part without explicit permission from the author(s).