The accuracy is extremely low... even after changing a lot of options of Method... Other methods' result is fine and normal, 0.80 accuracy like LogisticRegression. NetTrain works well

I've tried several hours in Linux/Max/Windows version of Mathematica 11.3

Where goes wrong?

resource = ResourceObject["MNIST"];
trainingData = ResourceData[resource, "TrainingData"];
testData = ResourceData[resource, "TestData"];
trainSet = RandomSample[trainingData, 20000];
model1 = Classify[trainSet, Method -> "NeuralNetwork"];
cm = ClassifierMeasurements[model1, testData];

enter image description here

  • $\begingroup$ How long did you allow it to train? Did you try just a bare Classify without the Method? $\endgroup$
    – Carl Lange
    Aug 3 '18 at 14:27
  • $\begingroup$ @CarlLange Other methods' result is Fine, 0.80 accuracy like LogisticRegression. TrainingTime is not the problem $\endgroup$ Aug 3 '18 at 15:19
  • $\begingroup$ Sorry, no idea. On my machine I get 0.9 accuracy with the same setup. By the way, please paste code into the question rather than a screenshot of the code. Nobody likes typing. $\endgroup$
    – Carl Lange
    Aug 3 '18 at 15:36
  • $\begingroup$ @CarlLange Code Added. What's your Mathematica Version? 11.3? $\endgroup$ Aug 3 '18 at 15:49
  • $\begingroup$ Yes, 11.3 on Mac. $\endgroup$
    – Carl Lange
    Aug 3 '18 at 16:04

The key problem is the EarlyStopping Option, I think the default setting of Automatic of EarlyStopping maybe True, This is not newbie-friendly.

enter image description here


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.