Suppose that you are training a network with parameters [4.5, 2.5, 1.2, 0.6] , a learning rate of 0.3, and a gradient of [-1, 9, 2, 5] . After one update step of gradient descent, what would your network’s parameters be equal to?

Which explanation is the mathematical reason for vanishing gradients?

Suppose that you are given a set of noisy data points of the height and speed of soccer players. What should you use to approximate the joint distribution over these variables?

Suppose that you are trying to fit a neural network into data that were sampled from a sine-curve function. Your network has only one input (phase). Which neural network is best suited for this?

Consider this matrix: With zero-padding=1 and a stride size of 2, what would be the result after min pooling with pooling size 2?

Which answer choice is the correct Linear regression assumption?

Suppose you are training a neural network. One of the neurons has an input vector [12, 0.5, 3, -10, 0.04], weights [1.3, -2.5, 4, 0.3, 1]. What is the input to this neuron? Give your answer to two decimal places.

Suppose that you want to synthesize images of a car by training a neural network. Which model should you use for this situation?

Identify and explain clearly whether any of the following scenarios have the hidden and/or exposed terminal problems. Explain. (Note: direction of arrows indicates information flow, but the signal propagates 360 degrees around transmitter; the existence of a dotted edge indicates that two nodes are within hearing range of each other)  

In CSMA/CA inter-frame spacings, explain why SIFS is shorter than DIFS. In a CSMA/CA timing diagram, does the DIFS interval come before or after the backoff interval? Why?