Vector Similarity Search at Scale // Dave Bergstein // MLOps Coffee Sessions #52
MLOps.community
00:00
Dog Versus Dog Verse Cat Verse Bird Verse Elephant
A neural net is trained to identify the difference between, was trained to identify dog verse cat verse bird verse elephant. And i break apart the neuronet, if you will, and i take the last fully connected layer. It's a thousand numbers that capture what makes zeus different than those other animals. If the numbers are close, then it's a good chance that the dog looks a lot like zoos. But if the numbers are distant, it's aGood chance the dog doesn't look like it, or maybe it's an elephant,. So we find i used cosin, no ti. Don't use tose ot product similarity. And i found the vectors closest, and they
Play episode from 12:24
Transcript
