Reducing avoidable bias

In the previous exercise, you realized that a Multi-Layer Perceptron (MLP) neural network with 3 hidden units was underfitting the pulsar_train dataset. In the video exercise, several options for solving the underfitting problem were mentioned, i.e., reducing the avoidable bias in your model. Be careful though: it is possible that while aiming to reduce bias, you end up increasing the variance.

In this exercise you are going to apply one of those options: increasing the model size. For your MLP model, this means increasing the number of hidden units in the hidden layer, which is controlled by the size parameter in the nnet() function call.

As before, the pulsar_train and pulsar_test datasets are available in your workspace and the nnet package has been pre-loaded.

Question

Which of the following is NOT an option to reduce model bias?

Possible Answers

Increase the model size.
Modify input features using error analysis.
Reduce or eliminate regularization.
Parallelize your algorithmic implementation.

Exercise

Reducing avoidable bias

Instructions 1/4

Question

Possible Answers