Anagnostopoulos, Georgios

Associate Professor
Electrical and Computer Engineering

Thank you for visiting my Faculty Profile page at FIT, which provides a snapshot of who I am and what I do here at Florida Tech. -GCA

Personal Overview

You can find more information about me on my personal web page, which is located at

http://my.fit.edu/~georgio

It includes pages to my ResearchMenteesInternal and External service, PresentationsShort BioPhotos, and Opportunities for students among a few other things. The last time I updated this profile page was on 18 Sepetember 2016.

Educational Background

Ph.D. in Electrical Engineering, University of Central Florida (UCF), 2001.
M.S. in Electrical Engineering. University of Central Florida, 1997.
Electrical Engineering Diploma, University of Patras (UoP), Greece, 1994.

Recognition & Awards

Florida Tech Shining Star (top 10 univeristy researcher), May 2017. [link]
Senior Member
, Institute of Electrical and Electronics Engineers (IEEE), November 2010.
Recipient, Kerry Bruce Clark Award for Excellence in Teaching, FIT, March 2007. [link]

Current Courses

Currently, I teach Signals & Systems (ECE 3222) in the Fall and Spring, Pattern Recognition (ECE 5258) in the Fall and Theory of Neural Networks (ECE 5268) in the Spring. Please refer to my Courses page for updated information.

Professional Experience

03/2008 to present

Associate Professor, Department of Electrical and Computer Engineering, Florida Institute of Technology, Melbourne, Florida.

01/2003 to 03/2008

Assistant Professor, Department of Electrical and Computer Engineering, Florida Institute of Technology, Melbourne, Florida.

01/2001 to 12/2002

Visiting Assistant Professor, Computer Science Department, University of Central Florida, Orlando, Florida.

08/2000
to 12/2001

Software Engineer, Lucent Technologies / Agere Systems, Orlando, Florida.

06/1997 
to 07/2000

Software Engineer, Technisource, Orlando, Florida.

 

In addition to these appointments, he was a...  

05/2012 to 06/2012

Visiting Researcher, US Air Force Research Laboratory, Rome, New York, USA.

10/2011 to 12/2011

Visiting Researcher, Algorithms & Complexity Group, Max Planck Institute for Informatics, Saarbrücken, Germany.

Additional Duties

Director, AMALTHEA REU Program, FIT; 2007 - present
Director, AEGIS RET Program, FIT; 2012 - present

Co-Director, Information Characterization & Exploitation (ICE) Laboratory, FIT; 2011 - present
Coordinator, Series of Electrical & Computer Engineering Seminars (SECES), FIT; 2009 - present
Member, Graduate Council, FIT; Fall 2014-present
Member, College Council, College of Engineering, FIT; 2009-present

Member, IEEE Computational Intelligence Society (CIS) Graduate Student Research Grants Committee (GSRGC), IEEE; February 2015 - present
Associate Editor, IEEE Transactions on Cybernetics (TCyb), IEEE; January 2015 – present

Current Research

Since of recent (2013), Dr. Anagnostopoulos' research has focused on the following:

  • Multi-Task Learning (MTL)
    Instead of learning a set of tasks independently of each other using meager individual data sets (which results in poor performance), MTL attempts to co-learn them (by pooling all task data) on the assumption that these tasks may be interrelated and that learning a task might benefit the learning of the remaining. MTL has found many applications in the current literature, especially in recommender systems. Dr. Anagnostopoulos and his doctoral students, Cong Li and Niloofar Yousefi, have worked on and continue pursuing the development of novel, kernel-based MTL frameworks with generalization performance guarantees. To date, their efforts have yielded 3 IEEE Transaction on Neural Networks & Learning Systems (TNNLS) papers and 2 ECML conference papers (ECML 2014, ECML 2015). On the subject of MTL, he gave a tutorial at the International Joint Conference on Neural Networks (IJCNN 2015) in Killarney, Ireland, in July 2015. Details of this tutorial can be found at multi-task-learning.info.
    Now, Dr. Cong Li is with Google, Inc. in Mountain View, CA.
     
  • Racing Algorithms (RAs)
    RAs are computational procedures for model selection under computational constraints. In specific, they allow for trading off between the likelihood of selecting the best model(s) out of a given ensemble with computational effort needed to make such distinctions. Hence, RAs play an important role in identifying the best performing model(s) for a given task. Dr. Anagnostopoulos and his doctoral student, Tiantian Zhang, have worked on and are still pursuing RAs. Zhang pioneered the first multi-objective RAs, S-Race and SPRITN-Race. Her 2013 paper on S-Race ended up as finalist for best paper award at the Genetic & Evolutionary Computation Conference (GECCO 2013). Zhang followed up her line of research with two more papers at GECCO, one of which was again nominated for best paper award (GECCO 2014). Her work on S-Race was published in IEEE Transactions on Cybernetics (TCyb), while her work on SPRINT-Race is pending publication in the same journal.
    As of recent, Dr. Tiantian Zhang joined Google, Inc. in Mountain View, CA.
     
  • Hash Learning
    Hash learning refers to learning a function that maps the original features of an object (e.g. audio/image/video) to a binary code (hash value) in order to achieve a specific task in a data-driven manner. A major application of hash learning is content-based information retrieval (CBIR); working with hash values (e.g. searching for similar objects to the query object) is computationally very efficient. Dr. Anagnostopoulos and his doctoral student, Yinjie Huang, have worked on and continue pursuing hash learning for CBIR based on mono- and multi-label information. So far, their work on hash learning via codewords will appear in a paper presented at ECML 2015.
    Dr. Yinjie Huang recently joined Twitter, Inc. in San Francisco, CA. 
     
  • Metric Learning
    Metric learning aims to learn the most suitable distance function (metric) for recognition tasks addressed via nearest-neighbor classification, so that similar (dissimilar) data samples are mapped closer to (apart from) each other. Dr. Anagnostopoulos and his doctoral student, Yinjie Huang, have developed a novel, kernel-based metric learning framework for classification that was presented at ECML 2013.
     
  • Stochastic Optimization Algorithms 
    Very often, in engineering applications (including some ML problems), cost functions to be minimized are ill-behaved: non-convex, multi-modal or non-differentiable. In such scenarios, classical numerical optimization techniques are practically non-applicable (e.g. fail to converge to a minimizer). Dr. Anagnostopoulos and his doctoral student, Azhar Khayrattee, have been focusing on swarm-based, derivative-free minimzation methods that can be utilized in such adverse contexts. One of their recent paper at GECCO 2014, which was nominated for best paper award, details such a method along with its convergence analysis.

 

Dr. Anagnostopoulos' recent (since 2010) research mentees were the following:

  • Joey Velez-Ginorio, Electrical Engineering junior, University of Central Florida, URE Participant in the ML 2 at UCF, Spring 2015 – present. 2017 Barry Goldwater Scholar [link]. The Barry Goldwater Scholarship award is the most prestigious undergraduate award in the US. He was one of the 250 recipients nation-wide and one of the 11 recipients from Florida.
  • Yinjie Huang, Ph.D. in Electrical Engineering, 2016, “Content-Based Information Retrieval Via Nearest- Neighbor Search,” ML 2 , Department of Electrical Engineering & Computer Science, University of Central Florida, 2010 – Summer 2016. Dr. Huang has joined Twitter, Inc. in San Francisco, CA.
  • Kunal Jagtap, M.S. in Computer Engineering, “Kernel Tangent Space Principal Component Analysis,” ICE Laboratory, Department of Electrical & Computer Engineering, Florida Institute of Technology, Spring 2016.
  • Yang Wang, M.S. in Electrical Engineering, “Document Classification Via Deep Auto-Encoding,” ICE Laboratory, Department of Electrical & Computer Engineering, Florida Institute of Technology, Spring 2015.
  • Haotian Chen, M.S. in Electrical Engineering, “Collapsing K-Means Clustering”, ICE Laboratory, Department of Electrical & Computer Engineering, Florida Institute of Technology, Spring 2015.
  • Tiantian Zhang, Ph.D. in Electrical Engineering, 2016, “Model Selection Via Racing,” ML 2 , Department of Electrical Engineering & Computer Science, University of Central Florida, 2010 – Spring 2016. Dr. Zhang has joined Google, Inc. in Mountain View, CA.
  • Cong Li, Ph.D. in Electrical Engineering, 2014, “On Kernel-Based Multi-Task Learning,” ML 2 , Depart- ment of Electrical Engineering & Computer Science, University of Central Florida, 2010 – Fall 2014. Dr. Li has joined Google, Inc. in Mountain View, CA.
  • Christopher Sentelle, Ph.D. in Electrical Engineering, 2014, “Practical Implementations of the Active Set Method for Support Vector Machine Training with Semi-definite Kernels,” ML 2 , Department of Electrical Engineering & Computer Science, University of Central Florida, 2008 – Spring 2014. Dr. Sentelle is a senior scientist for L3/Cyterra in Orlando, FL.
  • Naveed H. Iqbal, M.S. in Applied Mathematics, 2011, “Multinomial Squared Direction Cosines Regres- sion,” ICE Laboratory, Department of Mathematical Sciences, Florida Institute of Technology, 2009 – Summer 2011. Mr. Iqbal is pursuing a Ph.D. in Applied Mathematics at Florida Institute of Technology.
  • Rong Li, M.S. in Computer Engineering, 2011, “Multi-Objective Memetic Evolution of ART-based Classifiers,” Department of Electrical & Computer Engineering, Florida Institute of Technology, 2009 – Spring 2011. Mr. Li is a software engineer at SpectorSoft, Fort Pierce, FL.
  • Mingbo Ma, M.S. in Electrical Engineering, 2010, “Kernel-based Sammon Mapping for Dimensionality Reduction & Data Visualization,” Department of Electrical & Computer Engineering, Florida Institute of Technology, 2009 – Fall 2010. Mr. Ma is currently pursuing a Ph.D. in Computer Science at the City University of New York.
  • Assem Kaylani, Ph.D. in Computer Engineering, 2010, “An Adaptive Multi-Objective Evolutionary approach to optimize ARTMAP Neural Networks,” ML 2 , School of Electrical Engineering & Computer Science, University of Central Florida, (?) – Summer 2010. Dr. Kaylani is a development manager at InCube, Orlando, FL.

Selected Publications

 Journals and competitive conference publications since 2010:

  1. Tiantian Zhang, Michael Georgiopoulos, and Georgios C. Anagnostopoulos. Multi-objective model selection via racing. Cybernetics, IEEE Transactions on, 46(8):1863–1876, August 2016. Impact factor 4.943 [doi]
  2. Christopher G. Sentelle, Georgios C. Anagnostopoulos, and Michael Georgiopoulos. A simple method for solving the SVM regularization path for semidefinite kernels. IEEE Transactions on Neural Networks and Learning Systems, 27(4):709–722, April 2016. Impact factor 4.854 [doi]
  3. Yinjie Huang, Michael Georgiopoulos, Georgios C. Anagnostopoulos.  Hash Function Learning via Codewords. In Annalisa Appice, Pedro Pereira Rodrigues, Victor Santos Costa, Carlos Soares, Joo Gama, and Alpio Jorge, editors, Machine Learning and Knowledge Discovery in Databases, volume 9284 of Lecture Notes in Computer Science, pages 659–674. Springer International Publishing, 2015. Acceptance rate 23.4% (89/380) [doi]
  4. Niloofar Yousefi, Michael Georgiopoulos, and Georgios C. Anagnostopoulos. Multitask learning with group-specific feature space sharing. In Annalisa Appice, Pedro Pereira Rodrigues, Victor Santos Costa, Carlos Soares, Joo Gama, and Alpio Jorge, editors, Machine Learning and Knowledge Discovery in Databases, volume 9285 of Lecture Notes in Computer Science, pages 120–136. Springer International Publishing, 2015. Acceptance rate 23.4% (89/380) [doi]
  5. Cong Li, Michael Georgiopoulos, and Georgios C. Anagnostopoulos. Pareto-path multi-task multiple kernel learning. Neural Networks and Learning Systems, IEEE Transactions on, 26(1):51–61, Jan 2015. Impact factor 4.854 [doi]
  6. Cong Li, Michael Georgiopoulos, and Georgios C. Anagnostopoulos. Multitask classification hypothesis space with improved generalization bounds. Neural Networks and Learning Systems, IEEE Transactions on, 26(7):1468–1479, July 2015. Impact factor 4.854 [doi]
  7. Tiantian Zhang, Michael Georgiopoulos, and Georgios C. Anagnostopoulos. SPRINT multi-objective model racing. In Proceedings of the 2015 on Genetic and Evolutionary Computation Conference, GECCO ’15, pages 1383–1390, New York, NY, USA, 2015. ACM. Acceptance rate 36% (182/505) [doi]
  8. Cong Li, Michael Georgiopoulos, and Georgios C. Anagnostopoulos. Conic multi-task classification. In Toon Calders, Floriana Esposito, Eyke Hüllermeier, and Rosa Meo, editors, Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2014, Nancy, France, September 15-19, 2014. Proceedings, Part II, volume 8725 of Lecture Notes in Computer Science, pages 193–208. Springer, 2014. Acceptance rate 23.8% (115/483) [doi]
  9. Cong Li, Michael Georgiopoulos, and Georgios C. Anagnostopoulos. A unifying framework for typical multitask multiple kernel learning problems. Neural Networks and Learning Systems, IEEE Transactions on, 25(7):1287–1297, July 2014. Impact factor 4.854 [doi]
  10. Azhar Khayrattee and Georgios C. Anagnostopoulos. Derivative free optimization using a population-based stochastic gradient estimator. In Dirk V. Arnold, editor, Genetic and Evolutionary Computation Conference, (GECCO ’14), Vancouver, BC, Canada, July 12-16, 2014, pages 983–990. Association for Computing Machinery (ACM), 2014. Acceptance Rate: 33% (180/544) [nominee; Best Paper Award] [doi]
  11. Tiantian Zhang, Michael Georgiopoulos, and Georgios C. Anagnostopoulos. Online model racing based on extreme performance. In Dirk V. Arnold, editor, Genetic and Evolutionary Computation Conference, GECCO ’14, Vancouver, BC, Canada, July 12-16, 2014, pages 1351–1358. Association for Computing Machinery (ACM), 2014. Acceptance Rate: 33% (180/544) [nominee; Best Paper Award] [doi]
  12. Tiantian Zhang, Michael Georgiopoulos, and Georgios C. Anagnostopoulos. S-Race: A multi-objective racing algorithm. In Christian Blum and Enrique Alba, editors, Genetic & Evolutionary Computation Conference (GECCO), pages 1565–1572. Association for Computing Machinery (ACM), 2013. Acceptance rate 36% (204/570) [finalist; Best Paper Award] [doi]
  13. Cong Li, Michael Georgiopoulos, and Georgios C. Anagnostopoulos. Kernel-based distance metric learning in the output space. In Proceedings of the IEEE-INNS-ENNS International Joint Conference on Neural Networks (IJCNN), pages 1–8. Institute of Electrical and Electronics Engineers (IEEE), August 2013. Acceptance rate 72% (435/605) [nominee; Best Paper Award] [doi]
  14. Yinjie Huang, Michael Georgiopoulos, and Georgios C. Anagnostopoulos. Reduced-rank local distance metric learning. In Hendrik Blockeel, Kristian Kersting, Siegfried Nijssen, and Filip Zelezn´ y, editors, European Conference on Machine Learning (ECML), volume 8190 of Lecture Notes in Computer Science, pages 224–239. Springer, 2013. Acceptance rate 25% (111/443) [doi]
  15. Christopher Sentelle, Georgios C. Anagnostopoulos, and Michael Georgiopoulos. Efficient revised simplex method for svm training. Neural Networks, IEEE Transactions on, 22(10):1650–1661, Oct. 2011. Impact factor 4.854 [doi]
  16. Assem Kaylani, Michael Georgiopoulos, Mansooreh Mollaghasemi, Georgios C. Anagnostopoulos, Christopher Sentelle, and Mingyu Zhong. An adaptive multiobjective approach to evolving art architectures. Neural Networks, IEEE Transactions on, 21(4):529–550, April 2010. [doi]
  17. Rong Li, Timothy R. Mersch, Oriana X. Wen, Assem Kaylani, and Georgios C. Anagnostopoulos. Multi-objective memetic evolution of art-based classifiers. In Evolutionary Computation (CEC), 2010 IEEE Congress on, pages 1–8. Institute of Electrical and Electronics Engineers (IEEE), July 2010. [nominee; Best Paper Award] [doi]

 

By May 2017, Dr. Anagnostopoulos' research outcomes have been featured in 24 peer-reviewed journal papers, 6 book chapters, and 49 conference papers, of which 7 were invited. Details on these publications can be found on his Publications page. At the time of writing, his H-Index is 19 and he has 471 citations since 2012 as reported by Google Scholar. You are invited to visit his research profiles on Google ScholarMicrosoft Academic Search and Research Gate.