obel «at» ucsc·edu
Google Scholar link: Oceane's Google Scholar Page
Oceane recently completed her Ph.D. in CRSS/SSRC. Her current interest lies in machine learning applied to systems. She previously worked on CAPES: A Computer Automated Performance Enhancement System and Inkpack: Drive theft-resistant system with Kenneth Chang, Professor Miller, and Professor Long.
Now she is working on Geomancy: Automated Performance Enhancement through Data Placement Optimization which will serve as part of her thesis. She developed Geomancy, a tool that automatically optimizes the placement of data within a distributed storage system by leveraging a neural network architecture that accurately forecasts future performance-based access metrics.
Additionally, she is working with Sinjoni Mukhopadhyay on WinnowML: Model-based optimized feature selection, a feature selection method for system modeling. This method ingests a neural network architecture provided by the user, examines the dataset, and automatically explores the subsets of features within a dataset to determine the best subset that optimizes accuracy and training time. This project will serve as the second part of Oceane's thesis.
She is also working with Caltech on an optimized packet routing system which will be the third part of her thesis.
She did her undergraduate at USC, majoring in computer engineering and computer science. During her undergraduate career, she conducted research with Stanford on how to use robotics to teach computer science to k-12 students. Additionally, she conducted research on smartwatches and voice recognition with USC.
|Oct 1, 2020||Oceane Bel, Kenneth Chang, Nathan Tallent, Dirk Duellman, Ethan L. Miller, Faisal Nawab, Darrell D. E. Long, Scalable High-Performance QoS] [Prediction and Grouping] [Storage QoS]|
|Jun 4, 2018||Oceane Bel, Kenneth Chang, Daniel Bittman, Hiroshi Isozaki, Darrell D. E. Long, Ethan L. Miller, Storage Class Memories] [Secure File and Storage Systems]|
|Nov 13, 2017||
Ethan L. Miller,
Darrell D. E. Long,
CAPES: Unsupervised Storage Performance Tuning Using Neural Network-Based Deep Reinforcement Learning,Supercomputing '17, November 2017. [Scalable High-Performance QoS] [Tracing and Benchmarking] [Ultra-Large Scale Storage]
Click here for a list of recent collaborators.