Design for Gaussian Process
sourish-cmi opened this issue · comments
Initiating the discussion for Gaussian Process regression
I am thinking about the design for GP Regression
could be performed something like this:
container = fit(Formula, data::DataFrame, modelClass::GaussianProcessRegression,MeanFunction::MeanZero,kernelClass::Exponential)
For example,
container = fit(y~x1+x2+x3,train_data,GaussianProcessRegression(),MeanZero(),Exponential())
- This could implement the default as MLE through
Optimize
using theGaussianProcess.jl
.
The Bayesian method can be implemented using the following way:
container = fit(meanFun::Formula, kernelFun::Formula, data::DataFrame, modelClass::GaussianProcessRegression,kernelClass::Exponential
,prior::Prior_Normal(),sim_size::Int64)
For example,
container = fit(y~x1+x2+x3,train_data,GaussianProcessRegression(),Exponential(),Prior_Cauchy(),10000)
- This could be implemented using the
HMC
orMCMC
using theTuring.jl
.