Perceptron Text Classifier

Simple perceptron text classifier used for my class presentation. The program implements a single layer perceptron, the simplest neural network. Each perceptron learns to distinguish a class from others and is fed input from .arff files(Atribute-Relation File Format). Note that the images shown below were made by me (Zamfir Alexandru) in Apple Keynote.

image

Why use the Single Layer Perceptron?

  • it is easy to implement, serves as an introduction to neural networks;
  • it makes use of parallelization, since each perceptron is independent from each other;
  • it has a simple learning rule;

Why not use the Single Layer Perceptron?

  • Can only perfectly learn linearly separable data;
  • Mixed (but generally good) results when learning non-linearly separable data;

Measuring Methods / Evaluation Metrics

In order to measure the performance of our program, we will be using the Confusion Matrix.

image

Relevant formulas:

  • Precision: Measures the proportion of true positive predictions among all positive predictions made by the model.
  • Accuracy: Measures the overall correctness of the model's predictions.
  • Recall (Sensitivity): Measures the proportion of true positive predictions among all actual positive instances.
  • Specificity: Measures the proportion of true negative predictions among all actual negative instances.

Results

image