4

Question 1

What is the issue of dimensionality?

Answer 1

As the number of features or dimensions grows, the amount of data grows exponentially

Question 2

How LeNet works?

Answer 2

1 Convolutional layer
2. Subsampling (Max pooling)
3. Convolutional
4. Subsampling (Max pooling)
5. FC
6. FC
7. Gausian connections
Used Sigmoid or tanh

Question 3

How AlexNet works?

Answer 3

1. Uses Convulational layers to learn various patterns. Early layers detect low-level features while deeper layers detect more abstract features
2. ReLU is used to intruduce non-linearity
3. Max Pooling to reduce computation and makes the network more robust to translations and distortions
4. Fully Connected Layers - these layers combine the features extracted by the convolutional layers to make the final classification decisions
5. Softmax
   Used ReLU
   Slit into 2 GPUs