Inferring learning rate from learn.recorder.plot()

I’m currently watching Deep Learning Part 1- Lecture 2 (
@28:40, Jeremy finds the learning_rate to be 3e-5 from the graph given by learn.recorder.plot(). How does he infer it from the graph? Surely I’m missing something really obvious here. Any help is appreciated. Thanks!
Here, is a screenshot:

3e-05 is (roughtly) the point that has the steepest continuous slope in the graph, that is why we pick it.

1 Like

Thank you for your quick reply. Okay…I now understand the reason behind picking this point. Can you please tell me how he got the value 3e-05 as we progress from 1e-05 along the x-axis to 1e-04?

Absolutely! Yes there that probably should have been a better choice instead, uou are correct. Mabye 7 or 8e-4

I think I got it. Thanks again for your help.