search

Riashat / Deep-Bayesian-Active-Learning

Code for Deep Bayesian Active Learning (ICML 2017)
112 stars 35 forks source link

Random subsampling in Var Ratios and BALD #3

Open raulsoutelo opened 6 years ago

raulsoutelo commented 6 years ago

Hi,

In Var Ratios and BALD acquisition functions you first randomly sample a subset (2000 points) of the pool set and then select the points that maximize either Var Ratios or BALD in this subset. Is there any reason for that? Is it only to minimize the acquisition time? Sorry, I have not found any reference to that in the paper.

Thanks!

Riashat commented 6 years ago

Hi,

Yes, we subsampled 2000 datapoints at first only because, otherwise it gets computationally expensive to do acquisition over all the data points.

Hope that answers your question.

Thanks, Riashat

On Thu, Sep 13, 2018 at 10:11 AM Raúl Soutelo Quintela < notifications@github.com> wrote:

Hi,

In Var Ratios and BALD acquisition functions you first randomly sample a subset (2000 points) of the pool set and then select the points that maximize either Var Ratios or BALD in this subset. Is there any reason for that? Is it only to minimize the acquisition time? Sorry, I have not found any reference to that in the paper.

Thanks!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/Riashat/Deep-Bayesian-Active-Learning/issues/3, or mute the thread https://github.com/notifications/unsubscribe-auth/AHJNe_QKiCtRquuqYGe48rHNMc_6SRx8ks5uameugaJpZM4WnVjr .