Masterproef T704 : Efficient algorithms for low rank multivariate polynomial evaluation

Begeleiding:

Informatie:	Karl Meerbergen
Promotoren:	Karl Meerbergen	Nick Vannieuwenhoven
Begeleider:	Nick Vannieuwenhoven

Onderzoeksgroep:

Numerieke Approximatie en Lineaire Algebra Groep

Context:

Multivariate polynomials that admit a low-rank decomposition as a linear combination of products of linear polynomials may offer great computational savings over unstructured polynomials, both in terms of time and memory.

Doel:

The goal of this thesis is to understand which multivariate polynomials admit such a short expression and whether this representation is unique.

Uitwerking:

A computational technique for investigating this question will be developed. In addition, a simple algorithm will be developed, e.g., in Tensorlab, for approximating any multivariate polynomial by a low-rank multivariate polynomial. A fast evaluation algorithm will also be developed and compared with the naive evaluation strategy.

Relevante literatuur:

Profiel:

Deze masterproef is voor 1 student.