Which description is wrong about the hyperparameter?

Question

Options

• A1-typerparameters are parameters that set values before the algorithm begins learning.
• BMost machine learning algorithms have hyperparameters.
• CHyperparameters cannot be modified
• DThe value of the hyperparameter is not learned by the algorithm itself.

• A
  3% (1)
• B
  13% (5)
• C
  77% (30)
• D
  8% (3)

Explanation

Option C is the wrong description because hyperparameters absolutely can be modified - that's precisely how practitioners tune models. You set them before training, but you can change them between training runs using techniques like grid search, random search, or Bayesian optimization.

• A is correct: Hyperparameters (e.g., learning rate, number of layers) are configured prior to training, unlike model parameters (weights) which are adjusted during training.
• B is correct: Nearly all ML algorithms have hyperparameters - regularization strength in linear models, k in k-nearest neighbors, tree depth in decision trees, etc.
• D is correct: The algorithm learns internal parameters (weights, biases) from data, but hyperparameters are supplied externally by the practitioner - the model never updates them on its own.

Memory tip: Think of hyperparameters as the settings on an oven before you bake - you choose the temperature and time yourself, the oven doesn't decide them, but you can always change them for the next batch.

No community discussion yet for this question.