Machine Learning
High bias can cause an algorithm to underfit? 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. High bias is the main cause of underfitting. Bias is error from a model's assumptions being too restrictive for the real pattern, so the model misses that pattern even on its training data, and its error is high on training and new data alike.
Example: a linear classifier on the four XOR points, (0, 0) and (1, 1) in one class, (0, 1) and (1, 0) in the other. No straight line separates them, so it gets at most 3 of 4 right, however long it trains. Adding the feature x₁x₂ makes them separable and removes the bias. Heavy regularization adds bias too: ridge with a huge λ pushes weights toward zero and predicts roughly the mean.
Mind the word: bias here means error from wrong assumptions, not the bias term b in w·x + b, nor unfair treatment of groups. Common causes are weak features, too simple a model, too much regularization or stopping training early. Fixes run the opposite way, and more data rarely helps.