CUDA-MI: A Fast Method for Estimating Mutual Information
Author Information
Author(s): Shi Haixiang, Schmidt Bertil, Liu Weiguo, Müller-Wittig Wolfgang
Primary Institution: School of Computer Engineering, Nanyang Technological University, Singapore
Hypothesis
Can we accelerate B-spline function based mutual information estimation using CUDA?
Conclusion
CUDA-MI achieves significant speedup over traditional methods for estimating mutual information.
Supporting Evidence
- CUDA-MI achieves speedups of up to 82 times compared to a multi-threaded implementation.
- The method is open-source and publicly available.
- It effectively infers gene regulatory networks from microarray data.
Takeaway
This study created a new computer program that helps scientists understand how genes interact by calculating their mutual information much faster using special computer hardware.
Methodology
The study used CUDA programming to implement a parallel algorithm for estimating mutual information from gene expression data.
Limitations
The method may require significant GPU memory for large datasets, necessitating partitioning of data.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website