Chapter 15 Quiz: Auto-encoders

Chapter 16: Representation Learning with Auto-encoders

Learning objectives

Test your understanding of auto-encoder architecture, reconstruction error, and denoising concepts

This quiz covers both the lecture material and lab exercises from Chapter 15.

Architecture: Input $\to$ Encoder $\to$ Bottleneck ( $d \ll n$ ) $\to$ Decoder $\to$ Output.
Loss: Reconstruction MSE: $L = |x - \hat{x}|^2$ . Self-supervised (no labels).
VAE: Encoder outputs $\mu, \sigma$ . Reparameterization: $z = \mu + \sigma \odot \epsilon$ . Loss = reconstruction + KL divergence.
DAE: Corrupted input $\to$ clean reconstruction. Learns robust features.
Anomaly Detection: Train on normal data; flag high reconstruction error as anomalous.

Goodfellow, I., Bengio, Y., Courville, A. (2016). Deep Learning, ch. 14. MIT Press.
Kingma, D.P., Welling, M. (2014). Auto-encoding variational Bayes. ICLR.
Vincent, P., Larochelle, H., Lajoie, I., Bengio, Y., Manzagol, P.-A. (2010). Stacked denoising autoencoders. J. Mach. Learn. Res. 11, 3371–3408.