Qdrant’s Post

Transcript

Well, a very simple idea behind the vector databases that it's just a storage of any vectorized data and you can vectorize any data, you can vectorize whatever you want. It can be a video, an image, it's all the matter of embedding model. And then you're vectorizing your data and putting it inside the vector database. This vectors, these points in the space are grouped together based on their similarity and there are. Similarity can be measured by different metrics, but for example it can be a metric of the cosine similarity. This one depends on angle between these factors. I don't wanna be like annoyingly mathematical or whatever, but it gives you 2 very important things which are very interesting. So the similar objects by some category, some features, some class, some idea, they are going to go together and different objects are going to stay apart and. By these two facts similarity In this similarity you can actually use it to solve so many problems of analysis.