Quiz Interview Questions and answers on Gradient Descent

Welcome to your Quiz Interview Questions and answers on Gradient Descent

Select all the correct statement from the following.

We can use Gradient Descent to train Machine learning algorithms.

Gradient descent minimizes the cost function.

We can use Gradient Descent to train Neural Networks.

Gradient Descent approach calculates the parameters iteratively.

Which of following are input to gradient descent or hyper-parameter for GD?

Learning Rate

Maximum number of iterations

Random initialization of theta

All of above

It is always a great idea to set smallest learning rate for gradient descent.

True

False

Select all the correct statement from the following for gradient descent.

A smaller learning rate makes the solution less computationally expensive.

A larger learning rate may skip the global minima.

A large learning rate will require comparatively more time to find the solution.

None of Above

Is random values of parameters or weight has any impact on solution using gradient descent

No, it does not have any impact

Yes, it have the impact

Which of the following statement is true for batch gradient descent?

Batch gradient descent computes over a batch or portion of training instances.

Batch gradient descent computes over all training instances.

All of above conditions are possible.

None of above condition is possible.

Select all the correct statement from the following for stochastic gradient descent.

Stochastic Gradient Descent uses random instances in the training data at every step.

SGD algorithm becomes fast since it has to compute small data at every iteration.

Reduction in the cost function is smooth in Stochastic Gradient Decent.

SGD algorithm can skip the local minima.

Select all the correct statement from the following for mini batch gradient descent.

Computes over a random batch of training set.

Cost function is smooth

Cost function always reduces with number of iterations

All of above

Leave a Reply Cancel reply