Skip to content

Instantly share code, notes, and snippets.

@galvanic
Last active August 11, 2016 00:00
Show Gist options
  • Select an option

  • Save galvanic/e185bd316de11ace5d62095ecc121aa4 to your computer and use it in GitHub Desktop.

Select an option

Save galvanic/e185bd316de11ace5d62095ecc121aa4 to your computer and use it in GitHub Desktop.
test to see Adaline classifier performance depending on gradient descent mini-batch size (here 10) - vs. full batch and stochastic (batch size of 1)
metrics                                                 AUC      FNR      FPR  error_test  error_train
classifier          attack   % poisoned GD method                                                     
adaline             empty    0.0        mini-batch     0.94     0.00     0.12        0.03         0.01
                                        stochastic     0.58     0.65     0.20        0.54         0.51
                                        batch          0.93     0.00     0.14        0.03         0.01
                             0.5        mini-batch     0.92     0.01     0.15        0.04         0.01
                                        stochastic     0.45     0.57     0.53        0.56         0.29
                                        batch          0.90     0.02     0.19        0.06         0.00
logistic regression empty    0.0        mini-batch     0.95     0.01     0.09        0.03         0.00
                                        stochastic     0.95     0.05     0.06        0.05         0.02
                                        batch          0.97     0.01     0.04        0.02         0.00
                             0.5        mini-batch     0.95     0.01     0.08        0.03         0.00
                                        stochastic     0.93     0.03     0.11        0.05         0.01
                                        batch          0.94     0.01     0.11        0.03         0.00
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment