We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.SC

Change to browse by:

cs

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Symbolic Computation

Title: Fast Derivatives for Multilinear Polynomials

Authors: Valeri Aronov
Abstract: The article considers linear functions of many (n) variables - multilinear polynomials (MP). The three-steps evaluation is presented that uses the minimal possible number of floating point operations for non-sparse MP at each step. The minimal number of additions is achieved in the algorithm for fast MP derivatives (FMPD) calculation. The cost of evaluating all first derivatives approaches to only 1/8 of MP evaluation with a growing number of variables. The FMPD algorithm structure exhibits similarity to the Fast Fourier Transformation (FFT) algorithm.
Comments: 7 pages, 3 figures
Subjects: Symbolic Computation (cs.SC)
Cite as: arXiv:1911.02235 [cs.SC]
  (or arXiv:1911.02235v1 [cs.SC] for this version)

Submission history

From: Valeri Aronov [view email]
[v1] Wed, 6 Nov 2019 07:31:41 GMT (1219kb,D)

Link back to: arXiv, form interface, contact.