|
Qdrant |
Default
Question-answering system with Similarity Learning and Quaterion Many problems in modern machine learning are approached as classification tasks. Some are the classification tasks by design, but others are artificially transformed into such. And when you try to apply an approach, which does not naturally fit your problem, you risk coming up with over-complicated or bulky solutions. In some cases, you would even get worse performance.
Imagine that you got a new task and decided to solve it with a good old classification approach.