Are results somewhat dependent on the machine?


#1

I try to follow the course on Youtube but I generally cannot get as good results as shown in the lectures. I run the code on my own computer with a GeForce GTX970. Can the same code generate different results on a different machine?


#2

You need to be little more precise.


#3

For example, when running the “Rossmann” code (lesson3-rossman.ipynb) i only reach exp_rmspe about 0.12 while Jeremy Howard reach about 0.09. As you can se below, I start to overfit when reaching about 0.12 even if I run the same code.

Screenshot_2018-05-09_11-11-59