I am working on embedding, i would like to know which if Mean Squared error (MSE) is better to make comparison between two embedding vector than Co-sinus similarity. In which situation use one or another
Mean squared error (MSE) is a metric for the difference between observed and actual target values, typically used for regression. There are no actual target values (aka, labels) in embeddings, thus MSE is not appropriate.
Cosine similarity can compare vectors so it is appropriate embeddings.