add examples of how model fitting plots *should not* look?

[fedarko]

Idea from Antonio.

[mortonjt]

That's a great idea! The easiest way to do this is to bump up the learning rate to 1 and see what those tensorboard plots look like.

@antgonza are you interested in contributing this to the readme?

[antgonza]

I can ... just looked at the output and it looks terrible, so good call. I also checked the 0.5 and the output also looks as bad. Any suggestion on a value that will make it border line?