gpflow / docs Goto Github PK

I'm working in an instance where I want to fit a GPR to some training data and then use the posterior as a prior for some new data. I've fit the GPR to the data and things look good on that end, but I am stuck trying to feed the model in as a prior. The documentation does include this link on "Updating Model with New Data" that looks like it could be helpful, but the stub is empty.

I'm new to GPFlow and can't see in the documentation if such a thing is even possible. If you can provide me with any assistance, I would be most grateful!

Hi all, thanks for your great work on GPflow!

The documentation for gpflow.models.GPR says:

"The log likelihood of this model is sometimes referred to as the ‘log marginal likelihood’, and is given by
.

I find this misleading. The function log_marginal_likelihood(self) computes the "actual" log marginal likelihood - I'd suggest removing "is sometimes referred to as the ‘log marginal likelihood’"?

After mu, var = m.predict_f(x_grid) in classification.ipynb, it should be
plt.fill_between(x_grid.flatten(), np.ravel(sigmoid(mu + 2 * np.sqrt(var))), np.ravel(sigmoid(mu - 2 * np.sqrt(var))), alpha=0.3, color="C0").
with def sigmoid(x): return 1/(1+np.exp(x))
instead of
plt.fill_between(x_grid.flatten(), np.ravel(mu + 2 * np.sqrt(var)), np.ravel(mu - 2 * np.sqrt(var)), alpha=0.3, color="C0").