Using ANN to classify digit pictures.
- data set:
- data in website
- data in this repo
- two features: symmetry and intensity
- raw features: 256 grayscale values.
Plot a scatter plot of two features for 10 labels.
Apply neural network of 1 hidden layer to classify 1 and 5. The features are: symmetry and average intensity. Use 3-fold cross-validation.
Apply two-layer neural network for classification of 1 and 5, using the raw features as input. Apply 3-fold cross-validation. Train and test the following structures [256, 6, 2, 1], [256, 3, 2, 1].plot the change of the in-sample error and test-set error for each iteration.
Apply neural network and SVM for classification for all 10 digits, using the raw features as input.