Neural Network Regression

Question

I had a question, For a given data set X with two classes {0,1}. If I train two separate neural networks NN0 and NN1 for each class 0 and 1 respectively. Can NN0 predict points in the dataset from class 1, even though it was trained on class 0?

Answer 1

In short, no. This isn't how neural networks, or machine learning in general works. You train your model to recognise both of the classes of your data and that one model can then be used to predict the class of data it hasn't seen.

This is a great overview of what neural networks are, done by someone very smart: https://youtu.be/aircAruvnKk

This is a quick and dirty machine learning example by me: https://grahamlyons.com/article/machine-learning-for-the-lazy-beginner

Answer 2

When you test a case from X with NN0, and it tells you "I'm a 0", that's the exact same thing as it telling you "I'm not a 1", because your class set has size two, like a boolean.

Therefor, your NN0 network predicts 1s as well as 0s, so does your NN1.