2013/05/16 by Gabriel Kronberger, Kronberger, Gabriel, Michael Kommenda +1
Computer Science · #Advanced Multi-Objective Optimization Algorithms #Evolutionary Algorithms and Applications #FOS: Computer and information sciences #Gaussian Processes and Bayesian Inference #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Metaheuristic Optimization Algorithms Research #Neural and Evolutionary Computing (cs.NE)
paper · pdf · doi:10.48550/arxiv.1305.3794
openalex publication_date 2013/05/16 · openalex created_date 2022/09/19 · openalex updated_date 2026/07/28
In this contribution we describe an approach to evolve composite covariance\nfunctions for Gaussian processes using genetic programming. A critical aspect\nof Gaussian processes and similar kernel-based models such as SVM is, that the\ncovariance function should be adapted to the modeled data. Frequently, the\nsquared exponential covariance function is used as a default. However, this can\nlead to a misspecified model, which does not fit the data well. In the proposed\napproach we use a grammar for the composition of covariance functions and\ngenetic programming to search over the space of sentences that can be derived\nfrom the grammar. We tested the proposed approach on synthetic data from\ntwo-dimensional test functions, and on the Mauna Loa CO2 time series. The\nresults show, that our approach is feasible, finding covariance functions that\nperform much better than a default covariance function. For the CO2 data set a\ncomposite covariance function is found, that matches the performance of a\nhand-tuned covariance function.\n