Terminology Clarification

Course Q&A Deep Learning Specialization Improving Deep Neural Networks: Hyperparameter tun
1

When creating a post, please add:

  • Week # 1
  • Description:
    1- Is the term “Train Set Error” similar to the loss function (i.e. MAE, MSE, cross-entropy… result that we et when we train the network in python?

2- Is the term “Dev Set error” the same as the “validation loss” when training the neural network in Python?

please note that I train my network with a dataset of 80% training and 20% dev set. Please find attached a screenshot of the value I need clarification on.

image
image2630×110 19.1 KB

Thank you for helping.

2

Both loss and metrics specify the quantity of interest. If you set the same value to both these parameters as part of model.compile, they’ll report the same scores during model training. One thing to note about measures like accuracy is that it reports 1 - error.

In general, we use metrics for assessing model performance. When validation_data is set to Dev set, the observed metrics correspond to dev set metric.

Use model.evaluate to score a model without changing weights.

1 Like