Question: During the first iteration of a neural network algorithm, we initialize link weights and biases with random values. Do you think that such randomization has
During the first iteration of a neural network algorithm, we initialize link weights and biases with random values. Do you think that such randomization has any effect on the predictive performance of the classifier? If we had run the same neural network at two different times with two different sets of initial (and randomly generated) values, should we expect to see different results? Why / why not?
Step by Step Solution
There are 3 Steps involved in it
Get step-by-step solutions from verified subject matter experts
