Evgeni Tsivtsivadze
I am research scientist at the Institute for Computing and Information Sciences (iCIS),
Machine Learning Group, Radboud University Nijmegen.
I am also affiliated with the Netherlands Organization for Applied Scientific Research.
Email: evgeni[at]science.ru.nl
Currently my main research interests are in machine learning. I am working on kernel-based (and Bayesian) methods for preference learning (ranking), regression, and classification tasks as well as applications in, among others, bioinformatics, automated reasoning, natural language processing, and information retrieval.
Projects
Learning2Reason. Funded by NWO (Ongoing). The general aim of this project is to develop machine learning algorithms suitable for mining large corpora of formally expressed knowledge that are available in the fields of formal mathematics and software verification.GENEUSS (GEne NEtworks Underlying Synaptic Signaling). Funded by NWO (Past). Within this project we develop and apply machine learning techniques to construct a dynamic model for the presynaptic gene network underlying short-term plasticity in neuronal synapses.
BioText. Funded by TEKES (Past). This project concearns information extraction for identifying protein-protein interactions stated in biomedical text. We combine structural natural language processing with machine learning methods to address the general and domain-specific challenges of information extraction in biomedical domain.
Publications (bibtex file)
Journal articles and book chapters
Molecular machines in the synapse: Overlapping protein sets control distinct steps in neurosecretion.
PLoS Computation Biology, 2012. [ bib | http ]
Premise selection for mathematics by corpus analysis and kernel methods.
Journal of Automated Reasoning, 2012. Conditionally accepted. [ bib | http ]
Learning intransitive reciprocal relations with kernel methods.
European Journal of Operational Research, 206(3):676-685, 2010. [ bib | http ]
Co-regularized least-squares for label ranking.
Preference Learning, pages 107-123, 2010. [ bib | http ]
Kernels for text analysis.
Advances of Computational Intelligence, 116:81-97, May 2008. [ bib | http ]
Locality kernels for sequential data and their applications to parse ranking.
Applied Intelligence, 31(1):81-88, 2009. [ bib | http ]
An efficient algorithm for learning to rank from preference graphs.
Machine Learning, 75(1):129-165, 2009. [ bib | http ]
Conference and workshop papers
Learning semantics for automated reasoning.
NIPS Workshop on Learning Semantics, 2011. [ bib | .pdf ]
Semantic graph kernels for automated reasoning.
SIAM International Conference on Data Mining, 2011. [ bib | .pdf ]
Multi-output ranking for automated reasoning.
International Conference on Knowledge Discovery and Information Retrieval, 2011. [ bib | .pdf ]
Learning2reason.
Calculemus/MKM, pages 298-300, 2011. [ bib | http ]
Sparse preference learning.
NIPS Workshop on Practical Application of Sparse Modeling: Open Issues and New Directions, 2010. [ bib | .pdf ]
Efficient remote homology detection.
Pattern Recognition in Bioinformatics, 2010. [ bib | .pdf ]
Kernel principal component ranking: Robust ranking on noisy data.
ECML/PKDD-Workshop on Preference Learning (PL-09), 2009. [ bib | .pdf ]
Identification of presynaptic gene clusters in synaptic signaling using functional data from genetic perturbation studies in hippocampal autapses.
2nd INCF Congress of Neuroinformatics, Frontiers in Neuroinformatics, 2009. [ bib | http ]
From ranking to intransitive preference learning: Rock-paper-scissors and beyond.
ECML/PKDD-Workshop on Preference Learning (PL-09), 2009. [ bib | .pdf ]
Leanring preferences with co-regularized least squares.
ECML/PKDD-Workshop on Preference Learning (PL-08), pages 55-62, 2008. [ bib | .pdf ]
A sparse regularized least-squares preference learning algorithm.
Scandinavian Conference on Artificial Intelligence (SCAI 2008), pages 76-83. IOS Press, 2008. [ bib | .pdf ]
Regularized least-squares for learning non-transitive preferences between strategies.
Finnish Artificial Intelligence Conference , pages 94-98, IOS Press, 2008. [ bib ]
Locality kernels for protein classification.
International Workshop on Algorithms in Bioinformatics, (WABI 2007), pages 2-11. Springer, 2007. [ bib ]
Learning to rank with pairwise regularized least-squares.
SIGIR Workshop on Learning to Rank for Information Retrieval, pages 27-33, 2007. [ bib | .pdf ]
Locality-convolution kernel and its application to dependency parse ranking.
IEA/AIE 2006, Annecy, France, volume 4031 of Lecture Notes in Computer Science, pages 610-618. Springer, 2006. [ bib | .pdf ]
Graph kernels versus graph representations: a case study in parse ranking.
ECML/PKDD'06 workshop on Mining and Learning with Graphs (MLG 2006), Berlin, Germany, 2006. [ bib | .pdf ]
Regularized least-squares for parse ranking.
Advances in Intelligent Data Analysis VI (IDA 2005), pages 464-474. Springer, 2005. [ bib | .pdf ]
Information extraction from biomedical text: The biotext project.
Conference on Human Language Technologies (HLT 2005), pages 131-136. 2005. [ bib | .pdf ]
Software
Probabilistic Preference Learner/Ranker - ProbRank. Joint work with Botond Cseke.
-
The algorithm can learn a ranking function based on pairwise comparison data, that is, data about the ranking function values is provided in terms of pairwise comparisons at the given locations. This is accomplished in two ways: a) Approximating the marginal likelihood using expectation propagation and carrying out maximum likelihood procedure on the hyper-parameters. In this case the square exponential covariance function is used.
b) Considering ranking as a regression with Gaussian noise and Gaussian processes prior, given the score differences.
You can download Matlab implementation of the probabilistic preference learning models described in "Kernel principal component ranking: Robust ranking on noisy data".
E-MaLeS 1.0 (3rd place in the FOF division of the CADE ATP System Competition). Joint work with Daniel Kuehlwein.
-
E-MaLeS 1.0 is an automated theorem prover which is based on E prover.
E-MaleS 1.0 uses E with different strategies than the standard auto mode. Furthermore it employs strategy splitting, e.g. it runs several strategies.
Note that this version is very CASC focused. You can download the code here.
-
Contains the code and the data used in the experiments of the paper "Multi-Output Ranking for Automated Reasoning"
You can download the code here.