From the course: Advanced RAG Applications with Vector Databases

Unlock the full course today

Join today to access over 24,700 courses taught by industry experts.

Solution: Find anomalies in your embeddings

Solution: Find anomalies in your embeddings

From the course: Advanced RAG Applications with Vector Databases

Start my 1-month free trial Buy for my team

Solution: Find anomalies in your embeddings

“

(upbeat music) - [Instructor] All right, welcome to the solution piece. There are quite a few anomalies that you can find. Here, we're going to see two of them. The solution I chose to go with was to take the vector database in its retriever format and just use that directly, bypassing the LLM. So let's look at four examples. Two that show good data quality and two that show an anomaly. So we have this example from before where we saw the Rottweiler, which is dog number five. And that is correct. Here, let's also search with an image. So if we encode the image of the cat, we should get cat_1 and we see that we do get cat_1 back here as the top results. Now let's look for a gray cat with long hair in a field. So that's this cat, that's cat_2. And when we run this, we actually see that dog_5, the Rottweiler, is actually the top results. And we probably need to get a different embedding model or fine tune our embedding model in order to get the right result here. Here, we see golden…

Contents