Deep Learning Quiz 3

Vector of all partial derivatives with respect to every parameter is known as ___

• derivative

• gradient

• gradient descent

• neural network

We always move in a direction ____ the gradient in order to update the parameters

• same as

• opposite to

• perpendicular to

• parallel to

The ___ technique helps to reuse gradients, thereby speeding up computation

• backpropagation

• early stopping

• memoization

• gradient descent

Learning rate determines the size of the steps to perform updates.

• True

• False

• Cannot be determined

Given 1000 training samples, batch size of 50 and 20 epochs, the number of updates in every epoch is ___

• 1000

• 50

• 20

• 10

Assuming vanilla GD is being used, given 1000 training samples and 20 epochs, the total number of updates that a model's parameters undergo during the entire training cycle is ___

• 1000

• 50

• 20

• 10

Given 1000 training samples and a batch size of 1, can you determine the variant of gradient descent being used?

• Mini batch gradient descent

• Stochastic gradient descent

• Batch gradient descent

• Cannot be determined

Between Batch GD and Stochastic GD, which optimization algorithm requires lesser training time?

• Batch GD

• Stochastic GD

How can we specify batch size in code?

• batch_size argument of compile method

• batch_size argument of evaluate method

• batch_size argument of predict method

• batch_size argument of fit method

For the same number of epochs, we run batch GD as well as stochastic GD. Which algorithm will converge to the solution quicker?

• Batch GD

• Stochastic GD