2015-03-23

ADiJaC – Automatic Differentiation of Java Classfiles

Prof. Emil Slusanschi, University Politehnica of Bucharest, Bucharest

15 Jun 2015, 17:00–18:30; Location: S4|10-314

In this talk we discuss the current design and implementation of ADiJaC, an automatic differentiation tool for Java classfiles. ADiJaC uses source code transformation to generate derivative codes in both the forward and the reverse modes of automatic differentiation. We describe the overall architecture of the tool, and present various details and examples for each of the two modes of differentiation. We emphasize the enhancements that have been made over previous versions of ADiJaC, and illustrate their influence on the generality of the tool and the performance of the generated derivative codes. The ADiJaC tool has been used to generate derivatives for a variety of problems, including real-world applications. We evaluate the performance of such codes and compare it to derivatives generated by Tapenade, a well-established automatic differentiation tool. Additionally, we present detailed performance analyses of other two complex real-world applications. Apart from being the only general-purpose automatic differentiation tool for Java bytecode, we argue that ADiJaC’s features and performance are comparable to those of similar mature tools for other programming languages such as C/C++ or Fortran.

Category: CE Seminar

Contact

Technische Universität Darmstadt

Graduate School CE
Dolivostraße 15
D-64293 Darmstadt

Phone+49 6151/16-24401
Fax -24404
OfficeS4|10-322

to assistants' office

Open BSc/MSc Theses

Show a list of open BSc/MSc topics at GSC CE.

 Print |  Impressum |  Sitemap |  Search |  Contact |  Privacy Policy
zum Seitenanfangzum Seitenanfang