Softmax regression is a method in machine learning which allows for the classification of an input into discrete classes. Unlike the commonly used logistic regression, which can only perform binary classifications, softmax allows for classification into any number of possible classes. In this post I walk through the construction of a basic softmax classifier using python and ipython notebook. I show how to conduct digit recognition on the MNIST database, but the code can be applied to any number of machine learning problems. Feel free to reuse this code wherever may be helpful!