Suppose that I'm have 2 different datasets of same domain and trained a model with d1 dataset. Can we generalize the model to predict d2 dataset ? Is it possible or not. For example consider Dataset D1, Dataset D2 with images of cats and dogs and model is trained on D1. Then can we use the model to predict D2.

Any Kind of reference is helpful

Thanks in Advance

  • $\begingroup$ This is basically the point of training a model in the first place. Why would you train a model on D1, where you already know the answer, if you could not apply it anywhere else? $\endgroup$ – Nuclear Hoagie May 5 at 14:13
  • $\begingroup$ I would like to know about the distributions what if the D1 and D2 distributions are different? Will the generalization concept of Deep learning still holds? $\endgroup$ – SS Varshini May 5 at 14:17
  • $\begingroup$ You can look for transfer learning. It can work (with reasonable limitations) even when the datasets have different distributions $\endgroup$ – Nikos M. May 5 at 15:04
  • $\begingroup$ @Nikos M Even if we use transfer learning we are supposed to train/fine tune model with dataset(D2). My question is even D1 and D2 images are same but distribution may or may not be same So can i use model for D2 prediction $\endgroup$ – SS Varshini May 5 at 15:18
  • $\begingroup$ Under some assumptions it is possible yes. Unfortunately I cannot state the asumptions needed. Trial and error can be a guide $\endgroup$ – Nikos M. May 5 at 16:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.