Abstract
Stochastic gradient (SG) is the most commonly used optimization technique for maximum likelihood based approaches to independent component analysis (ICA). It is in particular the default solver in public implementations of Infomax and variants. Motivated by experimental findings on electroencephalography (EEG) data, we report some caveats which can impact the results and interpretation of neuroscience findings. We investigate issues raised by controlling the step size in gradient updates combined with early stopping conditions, as well as initialization choices which can artificially generate biologically plausible brain sources, so called dipolar sources. We provide experimental evidence that pushing the convergence of Infomax using non stochastic solvers can reduce the number of highly dipolar components and provide a mathematical explanation of this fact. Results are presented on public EEG data.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
References
Amari, S.-I.: Natural gradient works efficiently in learning. Neural Comput. 10(2), 251–276 (1998)
Bell, A.J., Sejnowski, T.J.: An information-maximization approach to blind separation and blind deconvolution. Neural Comput. 7(6), 1129–1159 (1995)
Bertsekas, D.: Nonlinear Programming. Athena Scientific, Cambridge (1999)
Bottou, L.: Large-scale machine learning with stochastic gradient descent. In: Lechevallier, Y., Saporta, G. (eds.) Proceedings of COMPSTAT’2010, pp. 177–186. Springer, Heidelberg (2010)
Cardoso, J.F.: Infomax and maximum likelihood for blind source separation. IEEE Signal Process. Lett. 4, 112–114 (1997)
Cardoso, J.F.: Blind signal separation: statistical principles. Proc. IEEE 86(10), 2009–2025 (1998)
Cardoso, J.F., Laheld, B.H.: Equivariant adaptive source separation. IEEE Trans. Sig. Proc. 44(12), 3017–3030 (1996)
Delorme, A., Makeig, S.: EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. J. Neurosci. Methods 134(1), 9–21 (2004)
Delorme, A., Palmer, J., Onton, J., Oostenveld, R., Makeig, S., et al.: Independent EEG sources are dipolar. PloS One 7(2), e30135 (2012)
Eldar, Y.C., Oppenheim, A.V.: MMSE whitening and subspace whitening. IEEE Trans. Inf. Theor. 49(7), 1846–1851 (2003)
Gramfort, A., et al.: MNE software for MEG and EEG data. Neuroimage 86, 446–460 (2014)
Jutten, C., Herault, J.: Blind separation of sources, Part I: an adaptive algorithm based on neuromimetic architecture. Signal Process. 24(1), 1–10 (1991)
Makeig, S., Bell, A.J., Jung, T.P., Sejnowski, T.J.: Independent component analysis of electroencephalographic data. In: NIPS, pp. 145–151 (1996)
Nunez, S.: Electric Fields of the Brain: The Neurophysics of EEG. Oxford University Press, New York (2006)
Onton, J., Delorme, A., Makeig, S.: Frontal midline EEG dynamics during working memory. Neuroimage 27(2), 341–356 (2005)
Pham, D.T., Garat, P.: Blind separation of mixture of independent sources through a quasi-maximum likelihood approach. IEEE Trans. Signal Process. 45(7), 1712–1725 (1997)
Schaul, T., Zhang, S., LeCun, Y.: No more pesky learning rates. In: Proceedings of ICML Conference (2013)
Scherg, M., Berg, P.: Use of prior knowledge in brain electromagnetic source analysis. Brain Topogr. 4(2), 143–150 (1991)
Acknowledgments
This work was funded by the Paris-Saclay Center for Data Science with partial support from the European Research Council (ERC-YStG-676943).
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2017 Springer International Publishing AG
About this paper
Cite this paper
Montoya-Martínez, J., Cardoso, JF., Gramfort, A. (2017). Caveats with Stochastic Gradient and Maximum Likelihood Based ICA for EEG. In: Tichavský, P., Babaie-Zadeh, M., Michel, O., Thirion-Moreau, N. (eds) Latent Variable Analysis and Signal Separation. LVA/ICA 2017. Lecture Notes in Computer Science(), vol 10169. Springer, Cham. https://doi.org/10.1007/978-3-319-53547-0_27
Download citation
DOI: https://doi.org/10.1007/978-3-319-53547-0_27
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-53546-3
Online ISBN: 978-3-319-53547-0
eBook Packages: Computer ScienceComputer Science (R0)