VIBE CODING QUIZ 2

What is the main job of a weight in a neural network?

it tells how strongly an input should affect the neuron

what is the main job of the bias in a neuron

it adds an adjustable offset to the weighted sum

why do neural networks use activation functions

to introduce nonlinearity so the network can learn more complex patterns

which expression correctly describes the value calculated before the activation function

z=w1×1+w2×2+…+b

what is ReLu (-2)

0

What is ReLu (3)

3

The output of a sigmoid activation function is normally between

0 and 1

The output of tanh is normally between

-1 and 1

a neuron has 100 input values. if it has one weight for each input and one bias, how many trainable parameters does it have

101

a neuron has x1=1, x2=2, w1=2, w2=1, and b=-1. what is z=w1×1+w2×2+b

3

if a weight is negative, increasing that input will usually

push the neuron’s weighted sum downward

what does forward propagation do

it uses current inputs, weights, biases, and activations to produce a prediction

what is the difference between loss and cost in these lecture notes

loss is for one example; cost is an average over several examples

using L=1/2(y_hat-y)², what is the loss when y_hat=0.6 and y=1.0

0.08

what does the learning rate control during gradient descent

the size of each parameter update step

activation functions help a neural network learn nonlinear relationships (TF)

true

the bias multiples every input before the weighted sum is calculated (TF)

false

ReLU passes positive input forward unchanged (TF)

true

gradient descent updates parameters by moving in the same direction as the gradient

false

if ReLU neuron’s pre-activation z is negative, its derivative is 0 for that example. (TF)

true