Detail publikace

Efficient Hardware Accelerator for Symbolic Regression Problems

Originální název

Efficient Hardware Accelerator for Symbolic Regression Problems

Anglický název

Efficient Hardware Accelerator for Symbolic Regression Problems

Jazyk

en

Originální abstrakt

In this paper, a new hardware architecture for the acceleration of symbolic regression problems using Cartesian Genetic Programming (CGP) is presented. In order to minimize the number of expensive memory accesses, a new algorithm is proposed. The search algorithm is implemented using PowerPC processor which is available in Xilinx FPGAs of Virtex family. A significant speedup of evolution is obtained in comparison with a highly optimized software implementation of CGP.

Anglický abstrakt

In this paper, a new hardware architecture for the acceleration of symbolic regression problems using Cartesian Genetic Programming (CGP) is presented. In order to minimize the number of expensive memory accesses, a new algorithm is proposed. The search algorithm is implemented using PowerPC processor which is available in Xilinx FPGAs of Virtex family. A significant speedup of evolution is obtained in comparison with a highly optimized software implementation of CGP.

BibTex


@inproceedings{BUT34289,
  author="Zdeněk {Vašíček} and Lukáš {Sekanina}",
  title="Efficient Hardware Accelerator for Symbolic Regression Problems",
  annote="In this paper, a new hardware architecture for the acceleration of symbolic
regression problems using Cartesian Genetic Programming (CGP) is presented. 
In order to minimize the number of expensive memory accesses, a new algorithm is
proposed.
The search algorithm is implemented using PowerPC processor which is available in
Xilinx FPGAs of Virtex family. 
A significant speedup of evolution is obtained in comparison with a highly
optimized software implementation of CGP.",
  address="Masaryk University",
  booktitle="5th Doctoral Workshop on Mathematical and Engineering Methods in Computer Science",
  chapter="34289",
  edition="NEUVEDEN",
  howpublished="print",
  institution="Masaryk University",
  year="2009",
  month="november",
  pages="192--199",
  publisher="Masaryk University",
  type="conference paper"
}