In this tutorial, we'll implement a Linear Classifier (i.e. Logistic Regression) to classify the MNIST data. You can read our step-by-step Tutorial on writing the code for this network, or skip it and see the implementation Code.
It's called linear classifier because it produces decision boundaries which are linear in nature. So, the classification made by this classifier is linear only. Figure below shows the structure of the classifier implemented for classifying the 28x28 MNIST data (i.e. images of hand-written digits).
Fig. 1- Structure of a linear classifier implemented for classifying MNIST digits
Example of decision boundaries generated for separating 3 classes are illustrated below.
Fig. 2- Decision boundries generate by a linear classifier
To learn more about the Logistic Regression, read this article.
- 1. Linear Classifier Implemented in TensorFlow
- 2. Linear Classifier+TensorBoard Visualizing the classifier structure, learning process and performance using TensorBoard