Java Deep Learning Cookbook

There's more...

Cosine similarity is the similarity between two nonzero vectors measured by the cosine of the angle between them. This metric measures the orientation instead of the magnitude because cosine similarity calculates the angle between document vectors instead of the word count. If the angle is zero, then the cosine value reaches 1, indicating that they are very similar. If the cosine similarity is near zero, then this indicates that there's less similarity between documents, and the document vectors will be orthogonal (perpendicular) to each other. Also, the documents that are dissimilar to each other will yield a negative cosine similarity. For such documents, cosine similarity can go up to -1, indicating an angle of 1,800 between ...