The at_gp from aboustati

at_gp's People

Contributors

Watchers

at_gp's Issues

About your code in final part

Hi Ayman,

Thanks for your sharing the code online. I am really like your code.
I am new to TensorFlow and GPflow. After reading your code, I am a bit confused about you final part of your code:

    Knn = self.kern.K(Xnew) if full_cov else self.kern.Kdiag(Xnew)
    f_mean, f_var = base_conditional(K_new, C, Knn, y, full_cov=full_cov, white=False)
    return f_mean + self.mean_function(Xnew), f_var

When I reading the paper: Adaptive Transfer Learning
http://www3.ntu.edu.sg/home/sinnopan/publications/[AAAI10]Adaptive%20Transfer%20Learning.pdf

the author may add some variance to the Knn in your code. I just slightly change your code to add the variance:

    Knn = self.kern.K(Xnew) + tf.eye(tf.shape(Xnew)[0], dtype=settings.float_type)\
               * self.likelihood.variance if full_cov else self.kern.Kdiag(Xnew)\
                                                           + self.source_likelihood.variance
    f_mean, f_var = base_conditional(K_new, C, Knn, y, full_cov=full_cov, white=False)
    return f_mean + self.mean_function(Xnew), f_var

I am not sure how the GPflow work in detail so I am not sure whether we need to add the variance in this way. Could you please clarify it?

Thanks in advance.

Recommend Projects