Cosine Similarity

A Simple Data Science Technique in the Era of Deep Learning

For Data Scientists and Software Engineers

The content would be useful to data scientists and software developers who need to support or produce systems that can compare and rank complex objects such as text documents, images, user profiles and so on.

On the one hand, Cosine Similarity is a simple technique that may or may not be always adequate. If one gets satisfactory performance, a lot of engineering and system maintenance costs could be saved.

On the other hand, the current trend is applying Machine Learning in the enterprise, is via Transfer Learning. In that case, Cosine Similariity, although simple, becomes very powerful, as one is able to make use of the knowledge available in complex text or image models, in a fairly simple way. This complex models are Deep Learning Models.