Solution - Building a Handwritten Digits Classifier

Just finished ‘building a handwritten digits classifier’ project. I took a different approach than one mentioned in the project. I instead used the cross_val_score function directly to see how I get my results. I’ve two notebooks with the exact same code one with scoring method as ‘neg_mean_squared_error’ and other ‘accuracy score’. Both of them give different conclusions I’m not sure why. I would really appreciate if someone can take a look at it.

Solution can be found at: github

Also, attached the solutionImageClassifier_scorer_accuracy.ipynb (128.6 KB) ImageClassifier_scorer_neg_mse.ipynb (125.5 KB)

Click here to view the jupyter notebook file in a new tab

1 Like

In the beginning, I also used MSE and I got a similar result as yours. Then I realized MSE calculates accuracy by comparing the distance between the actual values and our predictions, which does make sense with categorical labels. Perhaps MSE is inappropriate in this case.