Display options
Cite
Rachakonda S, Silva RF, Liu J, et al. Memory Efficient PCA Methods for Large Group ICA. Front Neurosci. 2016;10:17doi: 10.3389/fnins.2016.00017.
Rachakonda, S., Silva, R. F., Liu, J., & Calhoun, V. D. (2016). Memory Efficient PCA Methods for Large Group ICA. Frontiers in neuroscience, 1017. https://doi.org/10.3389/fnins.2016.00017
Rachakonda, Srinivas, et al. "Memory Efficient PCA Methods for Large Group ICA." Frontiers in neuroscience vol. 10 (2016): 17. doi: https://doi.org/10.3389/fnins.2016.00017
Rachakonda S, Silva RF, Liu J, Calhoun VD. Memory Efficient PCA Methods for Large Group ICA. Front Neurosci. 2016 Feb 02;10:17. doi: 10.3389/fnins.2016.00017. eCollection 2016. PMID: 26869874; PMCID: PMC4735350.
Copy Download .nbib Format: NLM
Share it on

Front Neurosci. 2016 Feb 02;10:17. doi: 10.3389/fnins.2016.00017. eCollection 2016.

Memory Efficient PCA Methods for Large Group ICA.

Frontiers in neuroscience

Srinivas Rachakonda, Rogers F Silva, Jingyu Liu, Vince D Calhoun

Affiliations

  1. The Mind Research Network and Lovelace Biomedical and Environmental Research Institute Albuquerque, NM, USA.
  2. The Mind Research Network and Lovelace Biomedical and Environmental Research InstituteAlbuquerque, NM, USA; Department of Electrical and Computer Engineering, The University of New MexicoAlbuquerque, NM, USA.
  3. The Mind Research Network and Lovelace Biomedical and Environmental Research InstituteAlbuquerque, NM, USA; Department of Electrical and Computer Engineering, The University of New MexicoAlbuquerque, NM, USA; Department of Computer Science, The University of New MexicoAlbuquerque, NM, USA.

PMID: 26869874 PMCID: PMC4735350 DOI: 10.3389/fnins.2016.00017

Abstract

Principal component analysis (PCA) is widely used for data reduction in group independent component analysis (ICA) of fMRI data. Commonly, group-level PCA of temporally concatenated datasets is computed prior to ICA of the group principal components. This work focuses on reducing very high dimensional temporally concatenated datasets into its group PCA space. Existing randomized PCA methods can determine the PCA subspace with minimal memory requirements and, thus, are ideal for solving large PCA problems. Since the number of dataloads is not typically optimized, we extend one of these methods to compute PCA of very large datasets with a minimal number of dataloads. This method is coined multi power iteration (MPOWIT). The key idea behind MPOWIT is to estimate a subspace larger than the desired one, while checking for convergence of only the smaller subset of interest. The number of iterations is reduced considerably (as well as the number of dataloads), accelerating convergence without loss of accuracy. More importantly, in the proposed implementation of MPOWIT, the memory required for successful recovery of the group principal components becomes independent of the number of subjects analyzed. Highly efficient subsampled eigenvalue decomposition techniques are also introduced, furnishing excellent PCA subspace approximations that can be used for intelligent initialization of randomized methods such as MPOWIT. Together, these developments enable efficient estimation of accurate principal components, as we illustrate by solving a 1600-subject group-level PCA of fMRI with standard acquisition parameters, on a regular desktop computer with only 4 GB RAM, in just a few hours. MPOWIT is also highly scalable and could realistically solve group-level PCA of fMRI on thousands of subjects, or more, using standard hardware, limited only by time, not memory. Also, the MPOWIT algorithm is highly parallelizable, which would enable fast, distributed implementations ideal for big data analysis. Implications to other methods such as expectation maximization PCA (EM PCA) are also presented. Based on our results, general recommendations for efficient application of PCA methods are given according to problem size and available computational resources. MPOWIT and all other methods discussed here are implemented and readily available in the open source GIFT software.

Keywords: EVD; PCA; SVD; big data; group ICA; memory; power iteration; subspace iteration

References

  1. Neural Comput. 1995 Nov;7(6):1129-59 - PubMed
  2. IEEE Trans Med Imaging. 2002 May;21(5):470-84 - PubMed
  3. IEEE Rev Biomed Eng. 2012;5:60-73 - PubMed
  4. Proc Natl Acad Sci U S A. 2010 Mar 9;107(10):4734-9 - PubMed
  5. Neuroimage. 2004 Jul;22(3):1214-22 - PubMed
  6. Proc Natl Acad Sci U S A. 2012 Feb 21;109(8):3131-6 - PubMed
  7. Hum Brain Mapp. 2001 Nov;14(3):140-51 - PubMed
  8. IEEE Trans Med Imaging. 2004 Feb;23(2):137-52 - PubMed
  9. Hum Brain Mapp. 2007 Nov;28(11):1251-66 - PubMed
  10. Front Syst Neurosci. 2011 Feb 04;5:2 - PubMed
  11. Magn Reson Med. 2012 May;67(5):1210-24 - PubMed
  12. PLoS One. 2010 Dec 20;5(12):e15710 - PubMed
  13. Front Hum Neurosci. 2013 May 01;7:168 - PubMed
  14. Magn Reson Imaging. 2006 Jun;24(5):591-6 - PubMed
  15. Neuroimage. 2014 Nov 1;101:738-49 - PubMed
  16. Neuroimage. 2015 Sep;118:662-6 - PubMed
  17. J Magn Reson. 2013 Apr;229:90-100 - PubMed
  18. Hum Brain Mapp. 2011 Dec;32(12):2075-95 - PubMed
  19. IEEE Trans Biomed Eng. 2011 Dec;58(12):3406-17 - PubMed

Publication Types

Grant support