Apache Spark Deep Learning Cookbook

There's more...

Ultimately, the goal is to get a predicted output that matches the actual output. Summing the product of the weights and the values helps achieve part of this process. Therefore, a random input of 0.5 and 0.5 would have a summation output of the following:

z = 0.5 * w1 + 0.5 * w2 + b 

Or it would have the following output with our current random values for our weights, w1 and w2:

z = 0.5 * (-0.2047) + 0.5 * (0.47894) + (-0.51943) = -7.557

The variable z is assigned as the product summation of the weights with the data points. Currently, the weights and biases are completely random. However, as mentioned earlier in the section, through a process called backpropagation, using gradient descent, the weights will be tweaked until ...