search

marionmari / pyGPs

pyGPs is a library containing an object-oriented python implementation for Gaussian Process (GP) regression and classification.
Other
213 stars 64 forks source link

not an issue..but a newbie question #47

Open shirishr opened 7 years ago

shirishr commented 7 years ago

How do I interpret result of model.predict(z, ys=np.ones((n,1)))? What are (ymu, ys2, fmu, fs2, lp)

shansfolder commented 7 years ago

Hi, @shirishr, the first four are the mean and variance of latent(f) and predictive variables(y). lp is the log predictive probability.

Furthermore, there are a few other properties you can get from the model object:

model.nlZ                   # negative log marginal likelihood
model.dnlZ.cov              # direvatives of negative log marginal likelihood
model.dnlZ.lik
model.dnlZ.mean
model.posterior.sW          # posterior structure
model.posterior.alpha
model.posterior.L
model.covfunc.hyp
model.meanfunc.hyp
model.likfunc.hyp
model.fm                    # latent mean
model.fs2                   # latent variance
model.ym                    # predictive mean
model.ys2                   # predictive variance
model.lp                    # log predictive probability

You may want to look at our very detailed documentation and tutorials: http://www-ai.cs.uni-dortmund.de/weblab/static/api_docs/pyGPs/index.html