Chapter 14 Quiz: Convolutional Neural Networks

Chapter 14: Convolutional Networks for Spatial Data

Learning objectives

Test your understanding of CNN architecture, convolution operations, pooling, and Keras implementation

This quiz covers both the lecture material and lab exercises from Chapter 14.

Convolution: Output size = $(n - f)/s + 1$ (valid padding). Same padding preserves dimensions.
Parameters: Conv2D with $F$ filters of size $k \times k$ on $C$ channels: $F \times (k^2 C + 1)$ params.
Pooling: Max/Avg pooling reduces spatial dimensions. $2 \times 2$ pool halves each dimension.
Weight Sharing: Same filter applied everywhere, reducing parameters and providing translation equivariance.
Transfer Learning: Pre-trained model (e.g., ImageNet) fine-tuned for new task.

Goodfellow, I., Bengio, Y., Courville, A. (2016). Deep Learning, ch. 9. MIT Press.
Krizhevsky, A., Sutskever, I., Hinton, G.E. (2012). ImageNet classification with deep convolutional neural networks. NeurIPS, 1097–1105.
Mousavi, S.M., Beroza, G.C. (2022). Deep-learning seismology. Science 377, eabm4470.