Which of the following statements best describes the concept of 'transfer learning' in the context of Convolutional Neural Networks for vision tasks?
Question 2
In a Convolutional Neural Network, what is the primary purpose of 'dropout'?
Question 3
Consider a convolutional layer with an input feature map of size $16 \times 16 \times 32$ (height, width, channels). If this layer uses 64 filters of size $3 \times 3$ with a stride of 1 and 'valid' padding (no padding), what will be the spatial dimensions (height and width) of the output feature map?
Question 4
What is the primary advantage of using 'depthwise separable convolutions' (as used in MobileNet architectures) compared to standard convolutions?
Question 5
In the context of modern CNN design, what is the main purpose of 'skip connections' or 'residual connections' (as popularized by ResNet)?
