Fast Genetic Programming on GPUs

Harding, Simon; Banzhaf, Wolfgang

doi:10.1007/978-3-540-71605-1_9

Cited by 86 publications

(66 citation statements)

References 17 publications

(15 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In previous work [Harding and Banzhaf, 2007] used the GPU exclusively for running training cases for cartesian genetic programming and showed impressive speed up in some cases but that improvement was highly variable. Indeed using the GPU was slower than the CPU in a few cases.…”

Section: Discussionmentioning

confidence: 99%

“…[Harding and Banzhaf, 2007, Section 3] described the various major high level language tools for programming GPUs (Sh, Brook, PyGPU and microsoft Accelerator). nVidia has two additional tools: CUDA and Cg (C for graphics [Fernando and Kilgard, 2003]).…”

Section: Programming Graphics Cardsmentioning

confidence: 99%

“…(See Table 3.) The task chosen was to predict the location of proteins within the cell given only their amino acid composition [Langdon and Banzhaf, -] [ Harding and Banzhaf, 2007]. A 1024 by 1024 population of programs of up to 63 tree elements and maximum depth of 8 was run on 200 of 1213 randomly chosen proteins selected for training.…”

Section: Evolving a Million Individuals For 1000 Generations Protein mentioning

confidence: 99%

“…It could be readily applied to cartesian GP [Harding and Banzhaf, 2007] and to linear GP. Typically both approaches use a small number of registers and do not require the use of a stack.…”

Section: Non-tree Gp Gp Without a Stackmentioning

confidence: 99%

“…Each stage being done by fixed specially hand written GPU programs. [Harding and Banzhaf, 2007] were able to show far higher (peak) speed ups when they ran the fitness evaluation of cartesian genetic programming on a GPU. [Chitty, 2007] used Cg to precompile tree GP programs on the host CPU before transferring them one at a time to a GPU for fitness evaluation.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

A SIMD Interpreter for Genetic Programming on GPU Graphics Cards

Langdon

Banzhaf

2008

Lecture Notes in Computer Science

Self Cite

104

View full text Add to dashboard Cite

Abstract. Mackey-Glass chaotic time series prediction and nuclear protein classification show the feasibility of evaluating genetic programming populations directly on parallel consumer gaming graphics processing units. Using a Linux KDE computer equipped with an nVidia GeForce 8800 GTX graphics processing unit card the C++ SPMD interpretter evolves programs at Giga GP operations per second (895 million GPops). We use the RapidMind general processing on GPU (GPGPU) framework to evaluate an entire population of a quarter of a million individual programs on a non-trivial problem in 4 seconds. An efficient reverse polish notation (RPN) tree based GP is given.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Programming Graphics Cardsmentioning

confidence: 99%

Section: Evolving a Million Individuals For 1000 Generations Protein mentioning

confidence: 99%

“…It could be readily applied to cartesian GP [Harding and Banzhaf, 2007] and to linear GP. Typically both approaches use a small number of registers and do not require the use of a stack.…”

Section: Non-tree Gp Gp Without a Stackmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

A SIMD Interpreter for Genetic Programming on GPU Graphics Cards

Langdon

Banzhaf

2008

Lecture Notes in Computer Science

Self Cite

104

View full text Add to dashboard Cite

show abstract

A graphical processing unit‐based parallel hybrid genetic algorithm for resource‐constrained multi‐project scheduling problem

Uysal

Sönmez

İşleyen

2021

Concurrency and Computation

View full text Add to dashboard Cite

In this article, we present a parallel graphical processing unit (GPU)-based genetic algorithm (GA) for solving the resource-constrained multi-project scheduling problem (RCMPSP). We assumed that activity pre-emption is not allowed. Problem is modeled in a portfolio of projects where precedence and resource constraints affect the portfolio duration. We also assume that the durations, availability of resources are deterministic and portfolio has a static nature. The objective in this article is to find a start time for each activity of the project so that the portfolio duration is minimized, while satisfying precedence relations and resource availabilities within a reasonable amount of time for small and large problem instances. In order to compare the efficiency of the proposed parallel GPU-based GA, problem is solved together with a CPU and a GPU.The results showed that GPU-based parallel GA has high potential for improving the performance of GAs for the RCMPSP particularly, for large-scale problems. K E Y W O R D Sparallel genetic algorithm, resource constrained multi-project scheduling problem, resource constrained project scheduling problem, resource management INTRODUCTIONThe objective of an optimization problem is to find the global optimal solution among all feasible solutions for a given search space. It is very common in engineering, management and other disciplines where an objective function defines optimality. Very often, mathematical models' complexity force to look for a good quality near optimal solution in a reasonable amount of time instead of finding the global optimal solution. Hence numerous research focus on developing meta-heuristics for finding good solutions in a short amount of computation time, in particular for complex problems.Metaheuristics are higher level heuristic methods which can be applied for different types of optimization problems. Among all meta-heuristics genetic algorithms (GAs) have extensive civil/construction engineering applications from structural optimization to scheduling. GAs are stochastic search methods which explores the search space while exploiting good solutions. 1 GAs depend on the randomness and use crossover and mutation to increase diversification. Solution of a problem can be represented as chromosomes and the best solution is selected according to fitness function calculations. Being a population-based method, GAs' performance highly depends on fitness calculations. GAs are effective in optimization problems but as problem size gets bigger several fitness calculations increase the computation time which reduces the efficiency of the algorithm. Fortunately, GAs are suitable for distributing the computational load to different cores which can increase the efficiency of the algorithm. 2 Therefore, there is a great potential to increase the efficiency of a GA by applying parallel computation methods.Early parallel computing attempts of a GA can be traced back to 1976. 3 Basic idea was to distribute the computational load among multiple computers within a lo...

show abstract