Câu 19: RAI: Risk and Artificial Intelligence

A quantitative analyst is studying the use of gradient descent method for training supervised models. Which of the following statements is most accurate regarding the gradient descent method?

1. A. The gradient descent method always finds the global minimum, even for high dimensional models.
2. B. In stochastic gradient descent, the hyperparameter is usually fixed before starting the iterations.
3. C. In the stochastic gradient descent method, the entire set of feature data is not needed in the memory at the same time. — đáp án hiện tại
4. D. The batch gradient descent method typically requires more iterations than the stochastic gradient descent method.