Data Science and Machine Learning (Theory and Projects) A to Z - Classical CNNs: AlexNet

The video introduces convolutional neural networks (CNNs), focusing on the historical context and significance of AlexNet, which won the 2012 ImageNet challenge. It compares the architectures of LeNet and AlexNet, highlighting their differences and similarities. LeNet is described with its use of average pooling and sigmoid activation, while AlexNet is noted for its deeper architecture, use of ReLU activation, and max pooling. The video emphasizes the shift from manually designed features to learned features in CNNs, marking a significant advancement in computer vision.

Who introduced the first convolutional neural network for handwritten digit recognition?

• Yann LeCun

• Andrew Ng

• John Lacoon

• Geoffrey Hinton

What is the size of the input image in LeNet?

• 32x32

• 28x28

• 224x224

• 64x64

In LeNet, what is the number of units in the final fully connected layer?

• 84

• 120

• 1000

• 10

How many convolutional layers does AlexNet have?

• 12

• 8

• 5

• 10

What activation function is used throughout AlexNet?

• Sigmoid

• Tanh

• Softmax

• ReLU

Which pooling method is consistently used in AlexNet?

• Average Pooling

• Max Pooling

• Global Pooling

• Min Pooling

What was a significant impact of AlexNet on the field of computer vision?

• It was the first network to use batch normalization.

• It was the first to use a fully connected layer.

• It showed that learned features can outperform manually designed features.

• It introduced the concept of dropout.