Hello,
I was just wondering that is there a way that i can use the trained gplvm model in pyro to reduce
dimensionality of similar datasets with same dimensions but different number of points? For example, i have a X_train and X_test set and i use X_train on the gplvm to reduce its dimensions and i was wondering if ican use the trained model to infer the DR for X_test? I can just use the whole X and reduce its dimensions and split into training and test but i was wondering if there is like a transform feature available like for PCA in sklearn?
Thanks!
Aneeq
@aneeqr Yes, you can with a little bit of effort. After training, your gplvm model learned base_model’s hyperparameters. You can create a new gplvm instance with the learned base_model (with fixed hyperparameters) and the new input. The result of running SVI/HMC on this new instance will be the one you need.
Thank you I will give it a go! I also wanted to ask that can you tell me how to use HMC on this model? I am a bit stuck on this part for sometime.
Thank you
You can run HMC on gplvm.model similar to any other model. However, I dont know if pyro’s hmc is stable enough to be able to do inference for gplvm.