GSGP-CUDA — A CUDA framework for Geometric Semantic Genetic Programming

Trujillo, Leonardo; Contreras, Jose Manuel Muñoz; Hernández, Daniel E.; Castelli, Mauro; Tapia, Juan J.

doi:10.1016/j.softx.2022.101085

Cited by 9 publications

(6 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While some of the GP-based methods, like Operon and GPlearn (https://gplearn.readthedocs.io/en/stable/index.html) (accessed on 31 January 2024), do allow for multithreaded processing, none of them were developed to exploit modern GPUs, the current standard of massively parallel devices. Moreover, while GP-based systems that run on GPUs are not new [26,42,43], they mostly implement basic GP-based algorithms that do not perform as well as the current state-of-the-art.…”

Section: Methodsmentioning

confidence: 99%

“…This section presents the core elements of M5GP, which as previously discussed, contains elements of both M3GP and M4GP, while utilizing some of the design choices used in the recently published CUDA-based GSGP-CUDA [26]. M5GP was developed using Python v3.9.12, using GPU processing with Numba kernels [53] and cuML [27] kernels for parameter estimation.…”

Section: Gpu-based Parallel Implementation Of M5gpmentioning

confidence: 99%

“…Individuals in M5GP, each one encoding a feature transformation, are encoded as fixed-length linear sequences of primitive elements, following [26]. Terminals include each of the original problem features (model inputs) x 1 , x 2 , • • •, x n , as well as random constants in the range [a, b].…”

Section: Individual Representation Interpretation and Parallel Proces...mentioning

confidence: 99%

“…After initialization, each individual is evaluated on each training sample (fitness case) in the training dataset. M5GP uses the same interpreter reported in [26]. Starting from the first gene or element in an individual, if it is a terminal element (feature or constant), it is pushed onto an output x ′ .…”

Section: Individual Representation Interpretation and Parallel Proces...mentioning

confidence: 99%

“…However, the proposal employs the solution representation of M4GP [23], along with a variation of the mutation operator for stack-based representations proposed in [25], which simplifies the encoding and search for multidimensional transformations of the input feature space. The proposed method also employs the GPU-based processing used in [26] for increased efficiency in the evolutionary process and parameter estimation of the resulting linear models methods with cuML [27]. M5GP is evaluated using the recently proposed benchmarking suite for SR called SRBench v2.0 [4].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

M5GP: Parallel Multidimensional Genetic Programming with Multidimensional Populations for Symbolic Regression

Cárdenas Florido,

Trujillo,

Hernandez

et al. 2024

MCA

Self Cite

View full text Add to dashboard Cite

Machine learning and artificial intelligence are growing in popularity thanks to their ability to produce models that exhibit unprecedented performance in domains that include computer vision, natural language processing and code generation. However, such models tend to be very large and complex and impossible to understand using traditional analysis or human scrutiny. Conversely, Symbolic Regression methods attempt to produce models that are relatively small and (potentially) human-readable. In this domain, Genetic Programming (GP) has proven to be a powerful search strategy that achieves state-of-the-art performance. This paper presents a new GP-based feature transformation method called M5GP, which is hybridized with multiple linear regression to produce linear models, implemented to exploit parallel processing on graphical processing units for efficient computation. M5GP is the most recent variant from a family of feature transformation methods (M2GP, M3GP and M4GP) that have proven to be powerful tools for both classification and regression tasks applied to tabular data. The proposed method was evaluated on SRBench v2.0, the current standard benchmarking suite for Symbolic Regression. Results show that M5GP achieves performance that is competitive with the state-of-the-art, achieving a top-three rank on the most difficult subset of black-box problems. Moreover, it achieves the lowest computation time when compared to other GP-based methods that have similar accuracy scores.

show abstract

Section: Methodsmentioning

confidence: 99%