Tutorial 102 - Deep Learning terminology explained - What is Training, Testing and Validation data

This video explains the primary difference between training, testing, and validation data.

Training data:
Data used to fit the model.
The model sees and learns from this data.
Accuracy gets checked and reported after each iteration.

Validation data:
Data used to evaluate a model during training.
Accuracy gets checked and reported after each epoch.
Model does not update any weights or biases (learn) based on this data.
The machine learning engineer monitors the validation error and tunes hyperparameters to optimize the model for better performance.

Test data:
Data used to evaluate the final trained model.
The machine learning engineer checks the accuracy after the entire training is done.
Often validation dataset is used as test dataset.
It is best practice to work with test data that represents generalized scenarios of future data.

[kongaaiguru]

Dear sir,
Thank you for your videos. I have a doubt while doing train and test split. I have data points of two categories 0 and 1. In the Excel sheet 500 points of category 0 (0, 500) are at one place and remain 500 data points of category 1 (501 to 1000)are at one place. If we split randomly during training will it take half of the data from category 0 and half of the data from category 1. Thank you sir

[abebefeyissa]

thank you for your video sir. but can you help me on Long term load forecasting using ANN?
thank you!

[PatriciaSantos-gtik]

If I have a small dataset and split it up to 70/30 would I do 30% for validation and 30% for testing? Or is it already 30% for validation only?