Menu:


Research Interests

My research broadly spans the areas of data science, signal processing, machine learning, and artificial intelligence. The overarching theme of my work has been the design and development of models and algorithms that can extract actionable and interpretable insights from multi-aspect/multi-modal data, typically with very little or no supervision. A major focus of my research has been on the development and advancement of tensor methods and their applications in high-impact real-world problems, including misinformation detection on the web, graph and social network analytics and mining, explainable AI, and detection of gravitational waves.

Download my CV

Quick links:

Contact

Email

  • epapalexcs dot ucr dot edu
  • vagelis.papalexakisgmail dot com

Location

3132 Multidisciplinary Research Building
Computer Science & Engineering Department
University of California Riverside
900 University Ave
Riverside, CA 92521
USA

Web Presence

The code provided in this page is provided as is. In case you spot a bug, please let me know. If you use some piece of code for your own work, please cite the corresponding article(s). Copyright and license as noted on each source file package.

AutoTen: Automatic Unsupervised Tensor Mining with Quality AssessmentNEW

Code for my SDM'16 paper that automatically decides the number of components for a PARAFAC decomposition in a data drive way, using the Core Consistency Diagnostic under various hypotheses for the data distribution.
Click here for the code (Requires the Tensor Toolbox for Matlab).

Fast and Efficient Core Consistency Diagnostic for Big Sparse Tensors

Source code for the corresponding ICASSP 2014 paper that introduces a scalable, fast, and efficient algorithm for computing the CORCONDIA diagnostic for the PARAFAC decomposition, which is essentially a measure of quality for the decomposition.
Click here for the code (Requires the Tensor Toolbox for Matlab).

ParCube: Sparse Parallelizable Tensor Decompositions

Fast, approximate and fully parallel algorithm that computes the PARAFAC decomposition. Matlab/Java code for the corresponding ECML-PKDD 2012 paper.
Click here for the code (Requires the Tensor Toolbox for Matlab).

UPDATE NOVEMBER 2014

Now the algorithm has full support of 4 mode tensors, as well as support for KL divergence in the core solver. Part of the code has been rewritten and further optimized.

Good-Enough Brain Model: Challenges, Algorithms and Discoveries in Multi-Subject Experiments

Demo and algorithm for the GeBM model introduced in the corresponding KDD 2014 paper.
Click here for the code.

Turbo-SMT: Accelerating Coupled Sparse Matrix-Tensor Factorizations by 200x

Fast, approximate and fully parallel algorithm that computes Coupled Matrix-Tensor factorizations. This is the implementation of our algorithm introduced in the corresponding SDM 2014 paper.
Click here for the code (Requires the Tensor Toolbox for Matlab and the CMTF Toolbox for Matlab).

GraphFuse: Tensor based multi-view Graph clustering

Matlab code for our tensor based technique for multi-view Graph clustering, as it appeared on our Fusion 2013 paper.
Click here for the code (Requires the Tensor Toolbox for Matlab).

PARAFAC with Sparse Latent Factors

Matlab code for our IEEE TSP 2013 paper, which introduces the PARAFAC decomposition with sparse latent factors, with application to Co-clustering.
Click here for the code (Requires the Tensor Toolbox for Matlab).

Co-clustering as a Decomposition with Sparse Latent Factors

Matlab code for our Co-clustering approach, as shown in our IEEE TSP 2013 paper, and our Journal of Chemometrics paper.
Click here for the code (Requires the N Way Toolbox for Matlab).