CODEWITHSUNDEEP
validationtestingneural-networkmse
user42423
user42423

Reputation: 117

plot mean square error for training validation and testing in neural network

Hi I see many of mean square error (MSE) plot as shown below:

enter image description here

My understanding to generate blue line (training) as below: Optimize the weights and bias values by using a certain optimization algorithm. For every epoch, calculate the MSE.

However, I don't understand how to generate the green and red lines.

Upvotes: 0

Views: 94

Answers (1)

Kurt Garcia
Kurt Garcia

Reputation: 44

Do you understand the purpose of the validation and test data?

Let's say you're training a neural network on 100,000 pieces of data. You should randomly split them into. Train(60,000), Validation(20,000), and Test(20,000).

The training data is what the network is actually trained out, the validation data is what checked against at each epoch: This is done to prevent overfitting on training data, and optimize hyperparameters.

At the very end, you can test the network against the test data, which is another test against overfitting.

Upvotes: 0

Related Questions