Computer Science Homework Help

Define Cost function

When is a cost function supposed to be large and when is it supposed to be small? When would it be 0?

Define Learning rate – include the purpose

Define Weights (aka learnable parameters) – include their purpose

a. What was the cost function for linear regression?

b. What was the cost function for logistic regression?

Which function(s) do we take the partial derivatives of and why?

Give a few examples of labelled data of one input variable for linear regression.

Give a few examples of y outputs for linear regression.

Give a few examples of labelled data of two input variables for logistic regression.

Give a few examples of y outputs for logistic regression (2-class classification).

Write psuedocode for the gradient descent algorithm.

What kind of plot might you look at as the gradient descent algorithm runs? What might you be looking for in this plot? Give two examples of plots that might not be ideal and what you would change and how in those two examples.

What’s the danger of too low a learning rate? What’s the danger of too high a learning rate?

If w is a column vector with the values w0, w1, w2, w3. What is w transpose w?

How can you use linear regression to learn a non-linear function?

undefined

How can you use logistic regression to learn a non-linear decision boundary?

undefined

What’s the idea behind regularization? What specific kind of regularization did we use for linear regression and logistic regression?

undefined

For a logistic regression algorithm to learn a linear decision boundary for data with 2 input variables and only 2 classes of that data — what actually gets learned to create the model?

undefined

How is that model then used to test unseen data?

undefined

If the gradient is convex, gradient descent will always work correctly and find the minimum. True or False. If false, when might it not?

undefined

If the gradient is non-convex, gradient descent will always work correctly and find the minimum. True or False. If false, when might it not?