Gaussian Process Regression (GPR) - Kriging
- originally in geostatistics, kriging or Kriging
- is a type of supervised regression method
- for a given set of training points, there are potentially infinitely many functions that fit the data. Gaussian processes offer an elegant solution to this problem by assigning a probability to each of these functions
- regression on Process
- is a method of interpolation based on the Gaussian process governed by prior covariances. Under suitable assumptions on the priors, GPR gives the best linear unbiased estimate (BLUE) at unsampled locations
- is a type of KDC) whose kernel is a Process (Bell Curve)
- can be thought of as an extension of the multivariate normal distribution to an infinite number of random variables covering each point on the input domain
GPR - Covariance/Kernel Function
GPR - Types of Models
Link to original
- Process (Bell Curve) - single-variable uni-modal Gaussian model
- Process (MVN) - multi-variable uni-modal Gaussian model
- Multimodal - multi-modal Gaussian model
- Process - multi-variable multi-model Gaussian model
GPR - Learning
- for learning Multivariate Unimodal Models: Gaussian Process Regression (GPR) - Explanation
- for learning Univariate Multimodal Models: EM - Gaussian Mixture Models
code examples:
GPR - Other
- Linear Regression vs Gaussian Regression
- Gaussian Process Regression vs Nadaraya-Watson Kernel Regression
Resources
- Gaussian Processes for Machine Learning ~ C. E. Rasmussen & C. K. I. Williams
- http://www.infinitecuriosity.org/vizgp/
- https://distill.pub/2019/visual-exploration-gaussian-processes/
- intuitive-tutorial-to-gaussian-process-regression.pdf