Interactive model training on tensorflow?

Question

Node.js

I can’t find an example of interactive training where data would be transferred to the model, based on them, it would make an assumption, and I would evaluate the result from 1 to 100.
For example, there are 3 levers. I submit some data array, the model selects the second lever, and I evaluate this choice at 5,
on the other array it selects the 1st lever and I evaluate it at 76.
I don’t look at examples, but they are all about image recognition and training on a prepared array data.
what would this code look like in nodejs? I’ll figure out where to get the data from and how to pass the input, but I can’t figure out how to evaluate the assumption that the model made ...

Answer 1

How tensorflow was made for machine vision, that's why pictures are everywhere. You need the nearest neighbors method - https://habr.com/en/post/149693/