Machine Learning
High variance can cause an algorithm to overfit. True or False?
True.
Card 1 of 259. Answer: True.
True.
Card 1 of 259. Answer: True.
About · Quizzes · History of AI · To suggest new questions or report an error send me a dm.
True. In the bias-variance sense, variance is how much a model's predictions change when it is trained on a different sample from the same population. A high-variance model bends to the noise in its particular training set, and fitting noise is what overfitting is. Strictly, high variance is less a cause of overfitting than its defining symptom.
Example: train an unpruned decision tree on five random samples of 1,000 homes and ask each to price the same house. The answers might range from $240,000 to $390,000, because each tree memorized different noise; a linear regression trained on the same samples might stay between $295,000 and $310,000.
Remedies that target variance include more data, bagging, regularization, fewer features and early stopping. On a learning curve, variance shows up as a train-validation gap that narrows as the training set grows. Do not confuse model variance with the variance of a feature or target: one describes how the fitted model shifts across samples, the other the data itself.