# Choosing a learning rate interval for linear separation with a perceptron

I'm having an issue with an exercise regarding the use of a perceptron for the linear separation of two sets M0 and M1. The question at hand is, how to go about finding an interval for a reliable learning rate of the perceptron.

The two sets are:

Or as a graph:

As can be seen from the graph, the x distance between the two sets is 1.

How would you go about finding an interval for a reliable learning rate? I'm guessing the lower limit for the learning rate would be close to zero but i'm having issues defining the maximum reliable learning rate without creating a model first and trying out different learning rates.