Interpretable deep neural networks for single-trial EEG classification

Abstract

Abstract-Background: In cognitive neuroscience the potential of Deep Neural Networks (DNNs) for solving complex classification tasks is yet to be fully exploited. The most limiting factor is that DNNs as notorious 'black boxes' do not provide insight into neurophysiological phenomena underlying a decision. Layerwise Relevance Propagation (LRP) has been introduced as a novel method to explain individual network decisions. New Method: We propose the application of DNNs with LRP for the first time for EEG data analysis. Through LRP the singletrial DNN decisions are transformed into heatmaps indicating each data point's relevance for the outcome of the decision. Results: DNN achieves classification accuracies comparable to those of CSP-LDA. In subjects with low performance subjectto-subject transfer of trained DNNs can improve the results. The single-trial LRP heatmaps reveal neurophysiologically plausible patterns, resembling CSP-derived scalp maps. Critically, while CSP patterns represent class-wise aggregated information, LRP heatmaps pinpoint neural patterns to single time points in single trials. Comparison with Existing Method(s):We compare the classification performance of DNNs to that of linear CSP-LDA on two data sets related to motor-imaginery BCI. Conclusion: We have demonstrated that DNN is a powerful nonlinear tool for EEG analysis. With LRP a new quality of highresolution assessment of neural activity can be reached. LRP is a potential remedy for the lack of interpretability of DNNs that has limited their utility in neuroscientific applications. The extreme specificity of the LRP-derived heatmaps opens up new avenues for investigating neural activity underlying complex perception or decision-related processes.

Machine learning for decoding listeners’ attention from electroencephalography evoked by continuous speech

“…Further, two cost functions were used to measure the distance between the estimated and the actual attended envelope: first, the MSE was used as a cost function C MSE . This is a widely used approach in deep learning (Bengio et al, 2007;Bengio, 2012;Sturm et al, 2016). With this cost function and the simple neural net, a linear regression is resembled by…”

“…While statistical models such as NNs are often considered to be black boxes, several methods exist to analyze the salient cues for classification learned by the net. An algorithm to analyze the relevance of the input features for producing the output (also referred to as heat mapping) was recently proposed (Bach et al, 2015;Sturm et al, 2016). In this study, the algorithm is used to identify where and when neural activity occurs that is relevant for decoding of auditory attention.…”

Interpretability of deep learning models: A survey of results

“…The output is a heatmap over the input features that indicates the relevance of each feature to the model output. This makes the method particularly well suited to analyzing image classifiers, though the method has also been adapted for text and electroencephalogram signal classification [31]. Samek et al [32] have also developed an objective metric for comparing the output of LRP with similar heatmapping algorithms.…”

Deep learning for situational understanding

“…The LRP technique has been used for EEG data analysis in [28]. Relevance score for each input data point is computed towards the final decision and is then visualized as a heat map providing interpretability.…”