2018/11/19 by Rahma Fourati, Fourati, Rahma, Boudour Ammar +5
Computer Science · Engineering · #Neural Networks and Reservoir Computing #Neural Networks and Applications #Advanced Memory and Neural Computing
paper · pdf · doi:10.48550/arxiv.1811.07516
In real-world applications such as emotion recognition from recorded brain\nactivity, data are captured from electrodes over time. These signals constitute\na multidimensional time series. In this paper, Echo State Network (ESN), a\nrecurrent neural network with a great success in time series prediction and\nclassification, is optimized with different neural plasticity rules for\nclassification of emotions based on electroencephalogram (EEG) time series.\nActually, the neural plasticity rules are a kind of unsupervised learning\nadapted for the reservoir, i.e. the hidden layer of ESN. More specifically, an\ninvestigation of Oja's rule, BCM rule and gaussian intrinsic plasticity rule\nwas carried out in the context of EEG-based emotion recognition. The study,\nalso, includes a comparison of the offline and online training of the ESN. When\ntesting on the well-known affective benchmark "DEAP dataset" which contains EEG\nsignals from 32 subjects, we find that pretraining ESN with gaussian intrinsic\nplasticity enhanced the classification accuracy and outperformed the results\nachieved with an ESN pretrained with synaptic plasticity. Four classification\nproblems were conducted in which the system complexity is increased and the\ndiscrimination is more challenging, i.e. inter-subject emotion discrimination.\nOur proposed method achieves higher performance over the state of the art\nmethods.\n