Python for Deep Learning - Build Neural Networks in Python - Gradient Descent versus Stochastic Gradient Descent

The video tutorial explains the concept of gradient descent, where the entire data set is used in each iteration to measure the gradient and update the cost function's parameters. It contrasts this with stochastic gradient descent, which uses a single value or a subset of values per iteration, making it faster for large data sets. The tutorial highlights the efficiency of stochastic gradient descent in handling large data sets compared to traditional gradient descent.

What is the primary method used in gradient descent to update parameters?

• Using a single data point

• Using a subset of data points

• Using the entire dataset

• Using random data points

How does stochastic gradient descent differ from gradient descent?

• It does not update parameters

• It is slower than gradient descent

• It uses a single value or subset of values for each iteration

• It uses the entire dataset for each iteration

Which method is generally faster when dealing with large datasets?

• Both are equally fast

• Stochastic gradient descent

• Neither, both are slow

• Gradient descent

What is a potential drawback of using gradient descent with large datasets?

• It may take too long

• It may use random data points

• It may not converge

• It may not update parameters

In which scenario would stochastic gradient descent be preferred over gradient descent?

• When the dataset is small

• When the dataset is sorted

• When the dataset is random

• When the dataset is large