 So, let's get to another cost function, cosensimilarity. Let us say we want two word embeddings to be close. Maybe we have a reason to believe that the words mean something similar. For example, because they're being used by the same scientist or in the same document. In that case, we might define cosensimilarity as minus x1, x2 divided by the two norms of those two vectors. Why is it called cosensimilarity? Well, the scalar product of x1 and x2 is the norm of x1 times the norm of x2 times the cosine of the angle between them. Just like a reminder, in high dimensional spaces, typical vectors have a 90 degree angle relative to one another. In any case, we want them to be similar. We can now use that on a natural language processing task. I should say we will be going over NLP later in this course. It's a very rich topic, but cosensimilarity is often used there. So the idea is I want two words that have similar meaning to be close to one another in some embedding space and words that have different meaning to be far away from one another. And it's time to run it. We prepared some text data for you. We want related words to be cosensimilar. We want unrelated words to be cosensimilar. Now, learn the meanings based on cosensimilarity.