Anastasios Kyrillidis (2014)

Research Interests:

  • Machine Learning
  • Convex and Non-convex Analysis and Optimization
  • Data Analytics and Mining
  • Structured Low Dimensional Models
  • Compressed Sensing

Biography

Anastasios Kyrillidis received his 5-year diploma and M.Sc. in Electronic and Computer Engineering from Technical University of Crete in 2008 and 2010, respectively. He was the first PhD student to graduate from the LIONS. Currently, he is Simons Foundation PostDoc Fellow at The University of Texas at Austin. His research interests includes convex and non-convex optimization, low-dimensional modeling in machine learning, and large-scale data analysis and processing. Anastasios was a PhD student at LIONS from September 2010 to October 2014. His PhD theses entitled Rigorous optimization recipes for sparse and low rank inverse problems with applications in data sciences was supervised by Professor Cevher.

Publications with LIONS (most recent)

A Non-Euclidean Gradient Descent Framework for Non-Convex Matrix Factorization

Y-P. Hsieh; Y-C. Kao; R. Karimi Mahabadi; Y. Alp; A. Kyrillidis et al. 

IEEE Transactions on Signal Processing. 2018. Vol. 66, num. 22, p. 5917-5926. DOI : 10.1109/TSP.2018.2870353.

A single-phase, proximal path-following framework

Q. Tran Dinh; A. Kyrillidis; V. Cevher 

Mathematics of Operations Research. 2018. Vol. 43, num. 4, p. 1326-1347. DOI : 10.1287/moor.2017.0907.

Convex block-sparse linear regression with expanders – provably

A. Kyrillidis; B. Bah; R. Hasheminezhad; Q. Tran Dinh; L. Baldassarre et al. 

2016. The 19th International Conference on Artificial Intelligence and Statistics (AISTATS 2016), Cadiz, Spain, May 7-11, 2016. p. 19-27.

Group-Sparse Model Selection: Hardness and Relaxations

L. Baldassarre; N. Bhan; V. Cevher; A. Kyrillidis 

IEEE Transactions on Information Theory. 2016. Vol. 62, num. 11, p. 6508-6534. DOI : 10.1109/TIT.2016.2602222.

Composite Self-Concordant Minimization

Q. Tran Dinh; A. Kyrillidis; V. Cevher 

Journal of Machine Learning Research. 2015. Vol. 16, p. 371-416.