InfoNCE Loss

#deep-learning #loss-in-ml #interview #need-review

InfoNCE loss is the improvement over the Contrastive Loss
In InfoNCE, there is one anchor, one positive and multiple negative samples for each instance
It converge faster than the Triplet Loss

InfoNCE Loss Formula

$l o s s = - E_{x} (l o g \frac{f (x^{+})}{f (x^{+}) + \sum_{i \in N} f (x_{i}^{-})})$

TODO:

Need to revise and understand the loss function
Especially the expected value part

Related Notes

Connected Pages

Depth

On this page

Pages mentioning this page

ML Interview

deep-learning

interview

loss-in-ml

Contrastive Learning