But What Is a Neural Network? | Deep Learning Chapter 1

1.How many pixels make up the input image that the neural network in this video is designed to recognize? 2 pts

• 28 × 28 = 784 pixels
• 16 × 16 = 256 pixels
• 10 × 10 = 100 pixels
• 32 × 32 = 1,024 pixels

2.In this neural network, the activation of a neuron is a number that can range from negative infinity to positive infinity. 2 pts

• True
• False

3.What is the purpose of the sigmoid function in a neural network? 2 pts

• It multiplies the weights of all connections together
• It squishes any real number into a value between 0 and 1
• It randomly selects which neurons will be activated
• It calculates the total number of parameters in the network

4.According to the video, approximately how many total weights and biases does the neural network described have? 2 pts

• 784
• 1,600
• 13,000
• 28,000

5.According to the video, the number of neurons in each hidden layer was chosen based on a precise mathematical formula. 2 pts

• True
• False

6.In the neural network described in the video, what does the output layer represent? 2 pts

• The grayscale value of each pixel in the image
• The weights connecting the last hidden layer to the output
• Ten neurons, each representing one of the digits 0–9
• The bias values added before the sigmoid function

7.The video suggests that modern neural networks commonly use the sigmoid function as their primary activation function. 2 pts

• True
• False

8.What is a 'bias' in a neural network, and why is it useful? Use details from the video in your answer. 4 pts

9.The video describes a 'best hope' for what the hidden layers of a neural network might be doing when recognizing handwritten digits. What is this hope, and do you think it is a reasonable way to design a system that can recognize patterns? Explain your thinking using examples from the video. 5 pts

10.Using information from the video, explain how a neural network is structured and how it processes information to recognize a handwritten digit. Your essay should cover the layers of the network, what neurons and activations are, how weights and biases work, and what the sigmoid function does. 10 pts