MathComputational mathOptimization

Pitfalls of gradient descent

Good steps

Report a typo

Given that you can compute both the objective function's gradient and directional derivative, what would be the approach to take in order to overcome the step size issue?

Select one option from the list

Apply the Wolfe conditions

Modify the step size by a constant factor

Use a random step size

___

Create a free account to access the full topic