FlexGP

Abstract

We describe FlexGP, the first Genetic Programming system to perform symbolic regression on large-scale datasets on the cloud via massive data-parallel ensemble learning. FlexGP provides a decentralized, fault tolerant parallelization framework that runs many copies of Multiple Regression Genetic Programming, a sophisticated symbolic regression algorithm, on the cloud. Each copy executes with a different sample of the data and different parameters. The framework can create a fused model or ensemble on demand as the individual GP learners are evolving. We demonstrate our framework by deploying 100 independent GP instances in a massive data-parallel manner to learn from a dataset composed of 515K exemplars and 90 features, and by generating a competitive fused model in less than 10 minutes.

Multiobjective genetic programming for reinforced concrete beam modeling

“…Instead of using one candidate, GP uses a group of individuals (known as population), formed by randomly combining mathematical building blocks such as constants, mathematical operators, analytic functions, state variables, and genetic operators, to make new individuals (generations) guided by fitness and complexity as objective functions that are meant to gauge the quality of each individual. The regression model 15,16 in Algorithm 1 (flowchart available in Reference 17) is implemented as an MOGP approach based on the work of Deb et al on nondominated sorting genetic algorithm II (NSGA-II). 18 Instead of fitness sharing (which was used in NSGA-I), NSGA-II uses the concept of crowding distance.…”

“…There are two values that the algorithm will minimize; first, the error, which can be the mean squared error (MSE) or mean absolute error (MAE), and second, the subtree complexity measure. 15,16,19 GP has great potential in predicting complex patterns. [20][21][22] It is also a suitable approach to be implemented using common parallelization frameworks such as message passing interface (MPI) and open multiprocessing (OpenMP).…”

Evolutionary computation: the next major transition of artificial intelligence?

“… EC has proven to work well in combination with many other AI techniques, including artificial neural networks [ 7 ] and other machine learning algorithms [ 8 ]. EC algorithms are inherently distributed, and are ripe for running in parallel on multi-core or distributed cloud-computing systems [ 9 ]. EC algorithms are anytime algorithms, meaning that they can provide a reasonable solution to a problem even when prematurely interrupted.…”

“…EC algorithms are inherently distributed, and are ripe for running in parallel on multi-core or distributed cloud-computing systems [ 9 ].…”

Parallel Bioinspired Algorithms on the Grid and Cloud

“…Finally we publish two papers based on the same ideas although with different implementations of the evolutionary algorithm [14]. The method is called Multiple Regression Genetic Programming (MRGP) which decouples and linearly combines a program's subexpressions via multiple regression on the target variable.…”