High performance reconfigurable computing for learning Bayesian networks with flexible parameterization
Abstract/Contents
- Abstract
- As improvements in high-throughput biology experiments outpace the advancement in processor performance, the gap between conventional processing capacities and new computational demands grows at exponential rates. In order to address this problem and keep pace with the new computing demands, traditional tool development approaches, where software and methods are developed independently of the hardware platform, need to be replaced by Hardware/Method co-development approaches. This new approach can achieve orders of magnitude higher performance as well as better power efficiency as the chip resources are customized to the particular computation demands of the problem. This work is an Algorithm/Architecture co-development effort to address a computationally intensive problem in modeling interaction networks using Bayesian statistics. Learning interaction networks from data uncovers causal relationships and allows scientists to predict and explain systems' behavior. Interaction networks have applications in many fields, though we will discuss them particularly in the field of personalized medicine where state of the art high throughput experiments generate extensive data on gene expression, DNA sequence and protein abundance. I will present how the FPGA-based HPC system we built, ParaLearn, is capable of modeling interaction networks in human T-Cells at orders of magnitude faster than conventional approaches.ParaLearn includes problem specific parallel/scalable algorithms, system software and hardware architecture to address this complex problem. ParaLearns accelerated and integrated solution enables scientists to use computationally intensive Bayesian network inference algorithms in real time settings. Also since it has been designed for high performance and flexibility, it creates new opportunities to develop network inference algorithms that are more robust to model parameterization and hence makes Bayesian network modeling a more useful and reliable tools for scientists.
Description
Type of resource | text |
---|---|
Form | electronic; electronic resource; remote |
Extent | 1 online resource. |
Publication date | 2010 |
Issuance | monographic |
Language | English |
Creators/Contributors
Associated with | Bani Asadi, Narges |
---|---|
Associated with | Stanford University, Department of Electrical Engineering |
Primary advisor | Wong, Wing Tak Jack Wong |
Thesis advisor | Wong, Wing Tak Jack Wong |
Thesis advisor | Meng, Teresa H |
Thesis advisor | Nolan, Garry P |
Advisor | Meng, Teresa H |
Advisor | Nolan, Garry P |
Subjects
Genre | Theses |
---|
Bibliographic information
Statement of responsibility | Narges Bani Asadi. |
---|---|
Note | Submitted to the Department of Electrical Engineering. |
Thesis | Thesis (Ph.D.)--Stanford University, 2010. |
Location | electronic resource |
Access conditions
- Copyright
- © 2010 by Narges Bani Asadi
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...