Skip to main content

Permuted MNIST MLP

Created on October 26|Last edited on October 26

Best results:


Showing first 10 runs
05101520Step0.950.9550.960.9650.970.9750.98
Showing first 10 runs
05101520Step0.10.150.20.25
Showing first 10 runs
05101520Step0.920.940.960.98
Showing first 10 runs
05101520Step0.050.10.150.20.25
Run set
184


Comparisions:



Run set
97

Observations are:
  • No. of neurons does not matter much. More neurons result in faster overfititng.
  • Smaller lr is always better, here lr < 0.1 is good.
  • Atleast 1 layer is needed.
  • No. of epochs depends on lr.