Deep Learning with PyTorch Quick Start Guide

Multiple features

In a realistic example, we would have more than one feature and each feature has an associated parameter value that requires fitting. We write the hypothesis function for multiple features as follows:

 

Here, x₀ is called the bias variable and is set to one, x₁ to x_n are the feature values, and n is the total number of features. Notice that we can write a vectorized version of the hypothesis function. Here, θ is the parameter vector and x is the feature vector. 

The cost function is still basically the same as the single feature case; we are just summing the error. We do, however, need to adjust the gradient descent rules ...