 This paper proposes a novel approach for detecting epileptic seizures from electroencephalogram, e.g., data. First, the raw e.g. signals are pre-processed using the wavelet transform to extract time-frequency distributions of the e.g. signals. Next, the data is then represented using a robust probabilistic collaborative representation, ProCRC, which combines the advantages of both kernel-based methods and graph-regularized non-negative matrix factorization, GNMF. This allows for better separation between seizure and non-seizure samples. Finally, the test sample is represented using ProCRC, and a decision tree algorithm is used to determine if the sample is a seizure or not. The proposed method was tested on the Freiburg e.g. database and achieved an average epic-based sensitivity of 96.48%, event-based sensitivity of 93.65% and specificity of 98.55%. This article was authored by Shasha Yuan, Jianwei Mu, Weidong Zhou, and others. We are article.tv, links in the description below.