Multiview Classification and Dimensionality Reduction of Scalp and Intracranial EEG Data through Tensor Factorisation

Abstract

Electroencephalography (EEG) signals arise as a mixture of various neural processes that occur in different spatial, frequency and temporal locations. In classification paradigms, algorithms are developed that can distinguish between these processes. In this work, we apply tensor factorisation to a set of EEG data from a group of epileptic patients and factorise the data into three modes; space, time and frequency with each mode containing a number of components or signatures. We train separate classifiers on various feature sets corresponding to complementary combinations of those modes and components and test the classification accuracy of each set. The relative influence on the classification accuracy of the respective spatial, temporal or frequency signatures can then be analysed and useful interpretations can be made. Additionaly, we show that through tensor factorisation we can perform dimensionality reduction by evaluating the classification performance with regards to the number mode components and by rejecting components with insignificant contribution to the classification accuracy.

Muscle Activity Analysis Using Higher-Order Tensor Decomposition: Application to Muscle Synergy Extraction

“…While one temporal component is assigned for each DoF since we do not need to segment main temporal activity. Thus, a [1,3,3] consTD is developed to estimate interpretable components from 1-DoF tensors, while a [2,5,5] model was used for the 2-DoFs tensors. [1,3,3] [2, 5, 5] g 1,n,n = 1 n ∈ {1, 2, 3} g 1,n,n = 1 n ∈ {1, 2, 5} g 2,n,n = 1 n ∈ {3, 4, 5} g = 0 otherwise g = 0 otherwise…”

“…The consTD models were applied on 1-and 2-DoFs tensors for muscle synergy estimation for 10 runs across the 27 subjects. The 1-DoF tensor was decomposed using [1,3,3] constrained Tucker method while the 2-DoFs tensor was decomposed using [2,5,5] constrained Tucker model. An example of [1,3,3] constrained Tucker method for Tensor shown in Figure 2e is illustrated in Figure 5.…”

“…Two consTD models ( [1,3,3] and [2,5,5]) were proposed to decompose the 1-DoF and 2-DoFs tensors respectively. They were able to achieve over 70% explained variance and decrease the execution time by about 10-fold compared to the non-negative Tucker model as shown in Table 4.…”

“…For instance, they have been utilised frequently in brain activity analysis [4]. Some applications include analysing electroencephalogram data to classify epileptic patients [5] and analysis of magnetoencephalogram activity in Alzheimer's disease [6]. Surprisingly, tensor factorisation had hardly been used in electromyography (EMG) analysis [7].…”

Auto-Weighted Multi-View Discriminative Metric Learning Method With Fisher Discriminative and Global Structure Constraints for Epilepsy EEG Signal Classification

“…Appropriately designed multi-view learning can significantly promote the performance of EEG signal classification. For example, Spyrou et al ( 2018 ) proposed a multiple features-based classifier to use spatial, temporal, or frequency EEG data. This classifier performs dimensionality reduction and rejects components through evaluating the classification performance.…”