Klaus-Robert Müller

TL;DR: This work proposes a novel technique, called sparsely connected sources analysis (SCSA), that can overcome the problem of volume conduction by modeling neural data innovatively with the following ingredients: the EEG/MEG is assumed to be a linear mixture of correlated sources following a multivariate autoregressive (MVAR) model.

TL;DR: In this article, a principled technique, Layer-wise Relevance Propagation (LRP), has been developed in order to better comprehend the inherent structured reasoning of complex nonlinear classification models such as Bag of Feature models or DNNs.

TL;DR: This work compares four popular neural network architectures, studies the effect of pretraining, evaluates the robustness of the considered alignment preprocessings via cross-method test set swapping and intuitively visualizes the model's prediction strategies in given preprocessing conditions using the recent Layer-wise Relevance Propagation (LRP) algorithm.

TL;DR: In this paper, the authors investigate three types of biases: biases which affect the entire dataset, biases which are by chance correlated with class labels and sampling biases, and they advocate pixel-wise heatmaps, which offer a more precise and versatile diagnostic instrument.

TL;DR: Deep SAD as discussed by the authors is an end-to-end deep methodology for general semi-supervised anomaly detection using an information-theoretic perspective on anomaly detection, which derives a loss motivated by the idea that the entropy of the latent distribution for normal data should be lower than that of the anomalous distribution.