Training and testing

5 important questions on Training and testing

Rule of thumb for training

Generalize rather than memorize

What is overfitting of a classifier in the process of training a supervised learning algorithm? What does this say about the quality of the classifier?

The performance of the classifier is much better on the training data than on the validation data. It implies the trained model does not properly generalize, also relies too much on memorization to perform well on its training data.

Why should negative data come from the same domain as the positive data?

The training will otherwise be irrelevant, because you would never see those instances in your application.
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

What are hyperparameters? How can they be optimized?

It are the parameters to configure the training algorithm. They can be optimized by assessing the performance of the trained model on the validation data.

What are the three steps of testing a supervised binary classifier algorithm?

  1. Extract the image descriptors of the images of the validation data set.
  2. Evaluate/Retrieve the labeling of the resulting images.
  3. Compare against the correct labels and obtain the accuracy

The question on the page originate from the summary of the following study material:

  • A unique study and practice tool
  • Never study anything twice again
  • Get the grades you hope for
  • 100% sure, 100% understanding
Remember faster, study better. Scientifically proven.
Trustpilot Logo