TSGP (Transfer Stacking Gaussian Process) is an algorithm that adaptively stacks pre-built gaussian process models from both the source and target domains in order to improve the predictive performance of the target regression problem.
TSGP takes as inputs (x_target,y_target,source_models), accepted inputs are shaped (n,d); n = # of instances, d = dimensions Accepted outputs are shaped (n,1), source_models are cells of RegressionGP models.
Example Use:
model = TSGP(x_target,y_target,source_models)
yhat = model.predict(x_test) Note: Scripts Tested in Matlab 2017b.
Tan, Alan, Ong, Y. S., Gupta, A., & Goh, C. K. (2017). Multi-Problem Surrogates: Transfer Evolutionary Multiobjective Optimization of Computationally Expensive Problems. IEEE Transactions on Evolutionary Computation.