Classifying Hand-Written Digits (Working Draft)

Hand-Written Digit Classification is an interesting problem and you might know why-

  1. This is an image classification problem.
  2. The search domain for this is pretty limited i.e. the digits can only range from 0-9.
  3. We have plenty of labelled datasets to test if our prediction is right or not.

So, let's see what are the various ways we could potentially solve this problem. We have plenty of ways of doing it - either we can employ classic machine learning methods to do this classification or we can build a time series classification model or employ neural networks to solve this kind of problem. Here is a blogpost on the comparative study of classical machine learning algorithms, namely-

  1. Linear Regression
  2. Logistic Regression
  3. K-Nearest Neighbours
  4. Decision Tree
  5. Support Vector Machines
  6. Random Forest
  7. Naive Bayes

Lastly, let us look at how would we do this using deep learning. We will be building a simple sequential classification model for it using a TensorFlow and Keras. One could also do this in PyTorch, which is a slightly better choice for quick model development and iteration.

Leave a Comment

Your email address will not be published. Required fields are marked *