Digits Dataset

Work on the digits dataset to classify handwritten digits using the provided dataset on Canvas using PLA.

ZipDigits.train is the training dataset and ZipDigits.test is the test dataset. The first column in the dataset

is the digit label, and the next $256$ columns are values between $-1$ and $1$ representing a grayscale image.

You will need to preprocess the data to get these labels. Take some time to familiarize yourself with the

dataset. We will work on this dataset throughout the course. Do the following preprocessing steps for both

training and testing datasets:

$$ Filter the dataset to include only digits labeled as $1$ and $5.$ Convert the labels into labels for binary

classification, i$.$e$.,1$ and $-1.$

$$ Extract intensity and symmetry features on the dataset as discussed in the class. You may use your

own mathematical definition of the two features. In the written part of the submission, include the

definition you used.

$$ For the training dataset, plot a $2$D scatter plot with the two features you extracted. Use different colors

and$/$or markers for the different classes. Submit the plot on the written part of your submission.

$$ Run PLA on train dataset to get optimal weights. Next, compute your accuracy on the test dataset

given the weights you found on training dataset