Speaker: Andrei Pătrașcu (University of Bucharest)

---

Abstract: We continue our adventure by investigating existing results using Gaussian Mixture Models (GMM) for anomaly detection and their adaptation to existing deep neural networks.

Required reading:

Zong, Bo, et al. “Deep autoencoding gaussian mixture model for unsupervised anomaly detection.” (2018).

Chapter 11 from Deisenroth, Marc Peter, A. Aldo Faisal, and Cheng Soon Ong. “Mathematics for Machine Learning.” (2018).

Speaker: Paul Irofti (University of Bucharest)

---

Abstract: We continue our investigation on the task of detecting outliers in networks when dealing with big-data and investigate existing online and distributed solutions.

Required reading:

Miao, Xuedan, et al. “Distributed online one-class support vector machine for anomaly detection over networks.” IEEE transactions on cybernetics 49.4 (2018): 1475-1488.

Liu, Zhaoting, Ying Liu, and Chunguang Li. “Distributed sparse recursive least-squares over networks.” IEEE Transactions on Signal Processing 62.6 (2014): 1386-1395.

Speaker: Andra Băltoiu (University of Bucharest)

---

Abstract: We continue our investigation on the task of detecting outliers in networks, by looking at the concept of signal variation on a graph.

Required reading:

A. Sandryhaila and J. M. F. Moura, “Classification via regularization on graphs,” 2013 IEEE Global Conference on Signal and Information Processing, Austin, TX, 2013, pp. 495-498.

S. Chen, A. Sandryhaila, J. M. F. Moura and J. Kovačević, “Signal Recovery on Graphs: Variation Minimization,” in IEEE Transactions on Signal Processing, vol. 63, no. 17, pp. 4609-4624, Sept.1, 2015.

Speaker: Andrei Pătrașcu (University of Bucharest)

---

Abstract: We continue our adventure by investigating existing results with Robust Principal Component Analysis (RPCA) and its adaptation to existing deep neural networks.

Required reading:

ZHOU, Chong; PAFFENROTH, Randy C. Anomaly detection with robust deep autoencoders. In: Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM, 2017. p. 665-674.

CANDÈS, Emmanuel J., et al. Robust principal component analysis?. Journal of the ACM (JACM), 2011, 58.3: 11.

Speaker: Andrei Pătrașcu (University of Bucharest)

---

Abstract: Recent empirical results confirm that one-class (OC) classification methods remain among the most important learning strategies for anomaly detection. In this seminar, we will technically describe in detail multiple basic OC schemes such as OC-SVM and SVDD and their deep variants, in order to identify room of improvements or generalization directions towards the graph anomaly detection context.