Institutional Repository [SANDBOX]
Technical University of Crete
EN  |  EL

Search

Browse

My Space

Efficient optimization algorithms for large tensor processing

Kolomvakis Christos

Full record


URI: http://purl.tuc.gr/dl/dias/C541BCFB-D331-44E8-BAF8-A1BE4A422678
Year 2021
Type of Item Master Thesis
License
Details
Bibliographic Citation Christos Kolomvakis, "Efficient optimization algorithms for large tensor processing", Master Thesis, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2021 https://doi.org/10.26233/heallink.tuc.90952
Appears in Collections

Summary

In this thesis, we consider the problem of tensor completion. We investigate two cases: In the first part, we consider Nonnegative Tensor Completion. We propose an improvement over an existing distributed algorithm for the solution of this problem, test it on synthetic and real datasets, and measure the execution time and speedups. In the second part, we consider unconstrained tensor completion with smoothing constraints. We present the problem statement and we propose a distributed algorithm for its solution. We develop an algorithm which takes into account the distribution of the nonzero elements during the assignment of subtensors (and, as a result, of the corresponding subfactors) to each processor. We test our adaptive partitioning algorithm on real world datasets and measure the attained speedup.

Available Files

Services

Statistics