Deep Learning Quiz 2

A McCulloch Pitts neuron takes 4 inputs and has a threshold of 3. If 3 inputs are high, what is the predicted output?

• 0

• 1

• 2

• 3

Select the point(s) of difference(s) between McCulloch Pitts (MP) neuron and Perceptron:

• MP neuron has weighted inputs while Perceptron inputs are unweighted

• Perceptron accepts real-valued input(s) while MP neuron accepts Boolean input(s)

• Perceptron uses the sigmoid activation function while MP neuron uses the step activation function

• Perceptron is able to learn non-linearities in the data while MP neuron is only able to capture linearities

The algorithm for training neural networks is ___

• Gradient descent

• Adam

• Backpropagation

• Forward propagation

Unsupervised Pre Training described an effective way to ___

• initialize weights of a deep neural network

• to detect traffic signs

• recognize handwritten digits

• recognize speech

LSTM networks have ___, ___ and ___ gates

To add a fully connected layer to a neural network, we use the ___ class in tensorflow

• FC

• Sequential

• Connect

• Dense

There are 1000 training samples and a batch size of 32 is chosen. Find the number of iterations per epoch

• 1000

• 100

• 1

• 32

We have 1000 training samples and 300 test samples. While calling the fit method on the model, we specify the validation_split = 0.3. If batch size is 10, find the number of batches through which training progresses

• 300

• 70

• 10

• 100

An MLP has an input layer with 4 features. The first hidden layer contains 8 neurons. The second hidden layer contains 4 neurons. The output layer contains 1 neuron. How many weight parameters are present?

• 68

• 64

• 32

• 4

An MLP has an input layer with 4 features. The first hidden layer contains 8 neurons. The second hidden layer contains 4 neurons. The output layer contains 1 neuron. How many bias parameters are present?

• 12

• 8

• 4

• 13