Pooya Saadatpanah
Hello! I'm Pooya Saadatpanah, a M.Sc student at the Department of Computer Science of the University of Toronto. I'm advised by Professor Marsha Chechik, and I'm working in the Software Engineering Group.
I graduated from Sharif University of Technology, Iran in 2011 with a Bachelor of Science in Computer Engineering.
Publications
Comparing the Effectiveness of Reasoning Formalisms for Partial Models [
pdf |
slides]
Pooya Saadatpanah,
Michalis Famelis,
Jan Gorzny,
Nathan Robinson, Marsha Chechik,
Rick Salay,
MoDeVVa 2012.
Uncertainty Management With Partial Models [
png] Michalis Famelis, Jan Gorzny,
Pooya Saadatpanah, Marsha Chechik and Rick Salay, presented in
RIA’12. (Poster Presentation)
Current Projects
Software Modelling in Presence of Uncertainty
Uncertainty is pervasive
in Model-based Software Engineering. Using partial models, modellers can perform
certain forms of reasoning, like checking properties, without having to prematurely
resolve uncertainty. In this project I design encoding of partial models in
SAT/SMT/CSP solvers. I have revisited current efficient model encodings and I
made them suitable for capturing all design alternatives and reasoning about them
in one query to reasoning engine. I have proposed a first-order encoding of partial
models in SMT solvers which enables unbounded property checking. Currently I'm
working on my encoding to make sure it scales well for extra large models.
Crowdsourced Data Cleaning and Entity Resolution
Combination of human
cognitive abilities and machines' computational powers seems like a strong alliance
for solving complex problems. In this project I'm designing an approach for data
cleansing across multiple connected database tables. The goal is to design a collective
entity resolution algorithm using crowdsourcing techniques. I propose an
efficient method to minimize the number of tasks humans should complete.
Past Projects
Gait Recognition Using Hidden Markov Model
In this project I addressed
motion recognition, the problem of identifying human motion using recorded
Mocap.
A single HMM is trained on sequences of two motion classes (walk and run), and two
distinct negative log likelihoods computed for each test sequence. A linear classier
is used to detect the type of motion.
Contact
Email: p
...@cs.toronto.edu
Facebook |
Linkedin
Address: 3270 - 40 St. George Street, Toronto, ON, Canada, M5S 2E4
View Larger Map
Last modified: October 2012
© 2012 Pooya Saadatpanah
Template design is shamelessly stolen from here