# Vortrag (TBA): Unterschied zwischen den Versionen

Gsc (Diskussion | Beiträge) |
Gsc (Diskussion | Beiträge) |
||

Zeile 2: | Zeile 2: | ||

|Titel DE=Multi-Objective Evolutionary Algorithm on GPGPU | |Titel DE=Multi-Objective Evolutionary Algorithm on GPGPU | ||

|Titel EN=Multi-Objective Evolutionary Algorithm on GPGPU | |Titel EN=Multi-Objective Evolutionary Algorithm on GPGPU | ||

− | |Beschreibung DE= | + | |Beschreibung DE=In the last two decades, the field of multi-objective optimization (MOO) has attracted researchers and practitioners to solve real world problems using evolutionary algorithms. For solving many-objective problems, a very large population is often required which takes time to both rank and evaluate. One of the remedies to reduce computation time is to perform evaluations in parallel. In recent years, an advent of massively parallel computing on general purpose graphic processing units (GPGPU) allows us to perform parallel computation on hundreds of threads. Now, a GPGPU card is an affordable commodity and can be used for scientific computing. In this talk, a GPGPU-compatible archive-based stochastic ranking evolutionary algorithm (G-ASREA) for MOO is discussed. Simulation results of G-ASREA show that it is approximately 5000 times faster than existing multi-objective evolutionary algorithms and approximately 15 times faster than serial versions of the archive-based stochastic ranking evolutionary algorithm. |

− | In the last two decades, the field of multi-objective optimization (MOO) has attracted researchers and practitioners to solve real world problems using evolutionary algorithms. For solving many-objective problems, a very large population is often required which takes time to both rank and evaluate. One of the remedies to reduce computation time is to perform evaluations in parallel. In recent years, an advent of massively parallel computing on general purpose graphic processing units (GPGPU) allows us to perform parallel computation on hundreds of threads. Now, a GPGPU card is an affordable commodity and can be used for scientific computing. In this talk, a GPGPU-compatible archive-based stochastic ranking evolutionary algorithm (G-ASREA) for MOO is discussed. Simulation results of G-ASREA show that it is approximately 5000 times faster than existing multi-objective evolutionary algorithms and approximately 15 times faster than serial versions of the archive-based stochastic ranking evolutionary algorithm. | + | |Beschreibung EN=In the last two decades, the field of multi-objective optimization (MOO) has attracted researchers and practitioners to solve real world problems using evolutionary algorithms. For solving many-objective problems, a very large population is often required which takes time to both rank and evaluate. One of the remedies to reduce computation time is to perform evaluations in parallel. In recent years, an advent of massively parallel computing on general purpose graphic processing units (GPGPU) allows us to perform parallel computation on hundreds of threads. Now, a GPGPU card is an affordable commodity and can be used for scientific computing. In this talk, a GPGPU-compatible archive-based stochastic ranking evolutionary algorithm (G-ASREA) for MOO is discussed. Simulation results of G-ASREA show that it is approximately 5000 times faster than existing multi-objective evolutionary algorithms and approximately 15 times faster than serial versions of the archive-based stochastic ranking evolutionary algorithm. |

− | |||

− | |Beschreibung EN= | ||

− | In the last two decades, the field of multi-objective optimization (MOO) has attracted researchers and practitioners to solve real world problems using evolutionary algorithms. For solving many-objective problems, a very large population is often required which takes time to both rank and evaluate. One of the remedies to reduce computation time is to perform evaluations in parallel. In recent years, an advent of massively parallel computing on general purpose graphic processing units (GPGPU) allows us to perform parallel computation on hundreds of threads. Now, a GPGPU card is an affordable commodity and can be used for scientific computing. In this talk, a GPGPU-compatible archive-based stochastic ranking evolutionary algorithm (G-ASREA) for MOO is discussed. Simulation results of G-ASREA show that it is approximately 5000 times faster than existing multi-objective evolutionary algorithms and approximately 15 times faster than serial versions of the archive-based stochastic ranking evolutionary algorithm. | ||

− | |||

|Veranstaltungsart=Kolloquium Angewandte Informatik | |Veranstaltungsart=Kolloquium Angewandte Informatik | ||

|Start=2013/06/07 14:00:00 | |Start=2013/06/07 14:00:00 | ||

Zeile 15: | Zeile 11: | ||

|Vortragender=Dr. Deepak Sharma | |Vortragender=Dr. Deepak Sharma | ||

|Eingeladen durch=Hartmut Schmeck | |Eingeladen durch=Hartmut Schmeck | ||

+ | |PDF=Vortrag Sharma.pdf | ||

|Forschungsgruppe=Effiziente Algorithmen | |Forschungsgruppe=Effiziente Algorithmen | ||

|In News anzeigen=True | |In News anzeigen=True | ||

}} | }} |

## Aktuelle Version vom 29. Mai 2013, 08:52 Uhr

## Multi-Objective Evolutionary Algorithm on GPGPU

##### Veranstaltungsart:

Kolloquium Angewandte Informatik

In the last two decades, the field of multi-objective optimization (MOO) has attracted researchers and practitioners to solve real world problems using evolutionary algorithms. For solving many-objective problems, a very large population is often required which takes time to both rank and evaluate. One of the remedies to reduce computation time is to perform evaluations in parallel. In recent years, an advent of massively parallel computing on general purpose graphic processing units (GPGPU) allows us to perform parallel computation on hundreds of threads. Now, a GPGPU card is an affordable commodity and can be used for scientific computing. In this talk, a GPGPU-compatible archive-based stochastic ranking evolutionary algorithm (G-ASREA) for MOO is discussed. Simulation results of G-ASREA show that it is approximately 5000 times faster than existing multi-objective evolutionary algorithms and approximately 15 times faster than serial versions of the archive-based stochastic ranking evolutionary algorithm.

(Dr. Deepak Sharma)

**Start:** 07. Juni 2013 um 14:00
**Ende:** 07. Juni 2013 um 15:30

**Im Gebäude** 11.40, **Raum:** 231

Veranstaltung vormerken: (iCal)

**Veranstalter:** Forschungsgruppe(n) Effiziente Algorithmen

Information: Media:Vortrag Sharma.pdf