Prediction
Explore how to use a trained perceptron model to make predictions for classification tasks. Understand the forward pass calculation, applying learned weights to new inputs, and how the model categorizes data points accurately.
We'll cover the following...
The ML process
For any complex problem that requires the computer to be able to identify patterns, there is an ML process to solve it.
This chapter demystifies each step of this process one by one. This lesson is about the fourth step—prediction.
Pattern identification using the model
The problem of categorizing movies as “Good” or “Bad” based on the quality of their direction and acting is a typical classification problem that can be solved using machine learning. We have learned how to visualize the patterns in the dataset and determine the optimal decision boundary in the input space that separates our movie data into two classes and also generalizes well. We call this optimal set of weights obtained after training the perceptron on the dataset of input-output pairs, our model, and it defines the optimal decision boundary.
With the trained model at hand, we can now easily solve the pattern identification problem, which was hard to do with simple logic-based programs. This means that the task of automatically identifying a movie as good or bad, given its direction and acting scores as input, becomes easy, given that the model is trained on a dataset of movies.
This trained model can now be deployed on our movies portal in real-time, and we can start classifying new movies as “Good” or “Bad.”
How does the model predict?
The question ...