0
$\begingroup$

Let $\mathcal{S}$ be the training input data set where each input $u^i \in \mathcal{S}$ has $d$ features. I want to design a ANN so that the cost function below is minimized (the sum of square of pairwise differences between model outputs) and the given constraint is satisfied, where $w$ is ANN model parameter vector.

enter image description here

Question: what kind of ANN is suitable for this purpose?[

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

A Siamese network (a network with multiple outputs) will work for such a case.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Why multiple outputs, could you please explain the reason? and What will be the number of outputs? $\endgroup$
    – data_science_learner
    May 2, 2021 at 21:51
  • 1
    $\begingroup$ Look at Triplet loss. Your question's has the same formulation like that. You would need three outputs. Then, you can reduce that loss. Or look at metric learning. There you get your embeddings for a batch and then try to reduce or increase distances between them. In case of metric learning you just need one output but a larger batch size will be needed. $\endgroup$
    – Abhishek Verma
    May 3, 2021 at 20:23

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.