About this Event
Electrical Engineering and Computer Science (EECS) Ph.D Dissertation Defense: Mahshid (Ashley) Montazer Qaem
"New Outliers Removal and Machine Learning Augmented Algorithms"
Mahshid (Ashley) Montazer Qaem
Electrical Engineering and Computer Science
University of California, Merced
This dissertation studies discrete optimization problems that resist optimum solutions because they are computationally hard or the inputs are revealed online, from two different angles. The first part of the dissertation revisits ranking and clustering problems in the presence of noise and outliers and shows how to effectively remove them to find more consistent solutions. The second part considers machine learning augmented algorithms in the online setting. Traditional research for online algorithms has mainly focused on their worst-case behavior. Unfortunately, the resulting algorithms don't often work the best in practice. There has recently been considerable effort to strengthen online algorithms by machine-learned predictions to achieve near optimum performance with good predictions while retaining some worst-case guarantees even with error-prone predictions. The dissertation makes further progress in this direction by revisiting classical scheduling and knapsack problems and proposing new types of predictions and prediction quality measures.
Mahshid (Ashley) Montazer Qaem is a Ph.D. candidate working with professor Sungjin Im in Electrical Engineering and Computer Science, University of California, Merced. She received her B.Sc. in Computer Science from University of Tehran in 2015. Her recent research interests include strengthening traditional worst-case algorithms with machine learned predictions. Her latest work in this direction has been selected as a best paper finalist at ACM Symposium on Parallelism in Algorithms and Architectures 2021. More broadly, she is interested in the design and analysis of algorithms, particularly in approximation and online algorithms.
1 person is interested in this event