Page 249 - Artificial Intelligence in the Age of Neural Networks and Brain Computing

240 CHAPTER 11 Deep Learning Approaches to Electrophysiological

also take advantage of engineered features derived from standard signal processing
methods, like frequency and time-frequency transforms. In recent years, the interna-
tional community has shown enormous interest in DL and artificial intelligence, by
funding several programs, in the public and private sectors. It is foreseen that these
programs will favor a relevant growth of economy and national gross value added.
However, in this chapter it has been shown that DL is just a contingent development
of ML and NN techniques originally proposed decades ago. This chapter focused on
a rather limited aspect of DL, namely the processing of multivariate time-series that
is relevant for biomedical applications but also pertinent to the future development
of IoT systems. The techniques here described can support a significant leap forward
in the real-time processing of unstructured data and in clinical diagnosis. Open
problems and limitations of DL have been discussed.

REFERENCES
[1] S. Haykin, Neural Networks, A comprehensive foundation. IEEE Press, 2004.
[2] K.I. Funahashi, On the approximate realization of continuous mappings by neural
networks, Neural Networks 2 (3) (1989) 183e192.
[3] Y. LeCun, Y. Bengio, G. Hinton, Deep learning, Nature 521 (7553) (2015) 436e444.
[4] G.E. Hinton, S. Osindero, Y.W. Teh, A fast learning algorithm for deep belief nets,
Neural Computation 18 (7) (2006) 1527e1554.
[5] G.E. Hinton, Training products of experts by minimizing contrastive divergence,
Training 14 (8) (2006).
[6] D. Erhan, Y. Bengio, A. Courville, P.A. Manzagol, P. Vincent, S. Bengio, Why does
unsupervised pre-training help deep learning? Journal of Machine Learning Research
11 (2) (2010) 625e660.
[7] A. Krizhevsky, I. Sutskever, G.E. Hinton, Image-net classification with deep convolu-
tional neural networks, in: Advances in Neural Information Processing Systems,
2012, pp. 1097e1105.
[8] V. Nair, G.E. Hinton, Rectified linear units improve restricted Boltzmann machines, in:
Proceedings of the 27th International Conference on Machine Learning (ICML-10),
2010, pp. 807e814.
[9] M.D. Zeiler, M. Ranzato, R. Monga, M. Mao, K. Yang, Q.V. Le, P. Nguyen, A. Senior,
V. Vanhoucke, J. Dean, et al., On rectified linear units for speech processing, in:
Acoustics, Speech and Signal Processing (ICASSP), 2013 IEEE International Confer-
ence on IEEE, 2013, pp. 3517e3521.
[10] P.L. Nunez, R. Srinivasan, Electric Fields of the Brain: The Neurophysics of EEG,
Oxford University Press, USA, 2006.
[11] S. Vaid, P. Singh, C. Kaur, EEG signal analysis for BCI interface: a review, in:
Advanced Computing Communication Technologies (ACCT), 2015 Fifth International
Conference on IEEE, 2015, pp. 143e147.
[12] G.R. Philips, J.J. Daly, J.C. Principe, Topographical measures of functional connectivity
as biomarkers for post-stroke motor recovery, Journal of Neuroengineering and
Rehabilitation 14 (1) (2017) 67.
[13] I.V.W.O. Tatum, Handbook of EEG Interpretation, Demos Medical Publishing, 2014.