Module 4

Question 1

Gradient descent

Answer 1

Repeatedly update parameters a and b by taking small steps in the negative direction of the partial derivative

Question 2

Gradient

Answer 2

Vector of all partial derivatives

Question 3

Logistic regression

Answer 3

• pass the regression output through a logistic function

- for binary classification

Question 4

Perceptron

Answer 4

• algorithm for supervised binary classification

- activation function: threshold function

Question 5

Linear activation

Answer 5

Directly passes on the output of the linear layer

Question 6

Sigmoid activation

Answer 6

Compresses the output smoothly into the range [0,1]

Question 7

Tanh activation

Answer 7

Compresses the output into the range [-1,1]

Question 8

ReLU activation

Answer 8

Rectified linear unit

• linear in the positive part
• non-linear overall

Question 9

Softmax activation

Answer 9

Scales the inputs into a probability distribution