When neural networks are employed for high-stakes decision-making, it is desirable that they provide explanations for their prediction in order for us to understand the features that have contributed to the decision. At the same time, it is important to flag potential outliers for in-depth verification by domain experts. In this work we propose to unify two differing aspects of explainability with outlier detection. We argue for a broader adoption of prototype-based student networks capable of providing an example-based explanation for their prediction and at the same time identify regions of similarity between the predicted sample and the examples. The examples are real prototypical cases sampled from the training set via a novel iterative prototype replacement algorithm. Furthermore, we propose to use the prototype similarity scores for identifying outliers. We compare performance in terms of the classification, explanation quality and outlier detection of our proposed network with baselines. We show that our prototype-based networks extending beyond similarity kernels deliver meaningful explanations and promising outlier detection results without compromising classification accuracy.

Anomaly detection algorithms find extensive use in various fields. This area of research has recently made great advances thanks to deep learning. A recent method, the deep Support Vector Data Description (deep SVDD), which is inspired by the classic kernel-based Support Vector Data Description (SVDD), is capable of simultaneously learning a feature representation of the data and a data-enclosing hypersphere. The method has shown promising results in both unsupervised and semi-supervised settings. However, deep SVDD suffers from hypersphere collapse-also known as mode collapse-, if the architecture of the model does not comply with certain architectural constraints, e.g. the removal of bias terms. These constraints limit the adaptability of the model and in some cases, may affect the model performance due to learning suboptimal features. In this work, we consider two regularizers to prevent hypersphere collapse in deep SVDD. The first regularizer is based on injecting random noise via the standard cross-entropy loss. The second regularizer penalizes the minibatch variance when it becomes too small. Moreover, we introduce an adaptive weighting scheme to control the amount of penalization between the SVDD loss and the respective regularizer. Our proposed regularized variants of deep SVDD show encouraging results and outperform a prominent state-of-the-art method on a setup where the anomalies have no apparent geometrical structure.

Recently conducted research demonstrated the potential use of mouse dynamics as a behavioral biometric for user authentication systems. However, the state-of-the-art methods in this field rely on classical machine learning methods that necessitate the design of hand crafted mouse features for feature extraction. To simplify the feature extraction process, we leverage various deep learning architectures for mouse movement sequences classification, including convolutional networks, recurrent networks, and a hybrid model which combines convolutional and recurrent layers. It is known that the training of these networks with random initialization of weights on small datasets will produce models that perform poorly. Therefore, we consider a two-dimensional convolutional neural network that allows transfer learning, which is a domain adaptation technique effective for learning on small datasets. Although employing such architecture may seem counterintuitive, since the temporal information is discarded from the input data, the architecture has outperformed all the other deep architectures investigated, as well as a classical machine learning method. In order to understand the features learned, we adopt the layer-wise relevance propagation (LRP) algorithm to compute relevance scores for each part of the mouse curves. In addition, the models are measured for their usability and effectiveness in realistic scenarios.