Machine learning and artificial intelligence have strong roots on principles of neural\ncomputation. Some examples are the structure of the first perceptron, inspired in the\nretina, neuroprosthetics based on ganglion cell recordings or Hopfield networks. In\naddition, machine learning provides a powerful set of tools to analyze neural data,\nwhich has already proved its efficacy in so distant fields of research as speech\nrecognition, behavioral states classification, or LFP recordings. However, despite the\nhuge technological advances in neural data reduction of dimensionality, pattern selection,\nand clustering during the last years, there has not been a proportional development of\nthe analytical tools used for Timeââ?¬â??Frequency (Tââ?¬â??F) analysis in neuroscience. Bearing this\nin mind, we introduce the convenience of using non-linear, non-stationary tools, EMD\nalgorithms in particular, for the transformation of the oscillatory neural data (EEG, EMG,\nspike oscillationsââ?¬Â¦) into the Tââ?¬â??F domain prior to its analysis with machine learning tools.\nWe support that to achieve meaningful conclusions, the transformed data we analyze\nhas to be as faithful as possible to the original recording, so that the transformations\nforced into the data due to restrictions in the Tââ?¬â??F computation are not extended to\nthe results of the machine learning analysis. Moreover, bioinspired computation such\nas brainââ?¬â??machine interface may be enriched from a more precise definition of neuronal\ncoding where non-linearities of the neuronal dynamics are considered.
Loading....