Generative Models for Deep Learning Engineers
Generative models learn a data distribution so they can produce new samples. Autoencoders and variational autoencoders compress data into latent codes, GANs train a generator against a discriminator, and diffusion and flow-matching models create samples by gradually transforming noise into data. Each family trades off sample quality, diversity and speed.
Stage: Core Architectures · Part of the Deep Learning Engineer roadmap · Last reviewed 19 November 2026
What to learn in Generative Models
- Train an autoencoder and visualise its latent space
- Write the VAE loss from reconstruction and KL terms
- Explain the generator and discriminator game in a GAN
- Describe the forward noising and reverse denoising steps of diffusion
- Compare sampling speed across VAEs, GANs and diffusion
- Explain classifier-free guidance for conditional generation
Generative Models interview questions
What is the reparameterisation trick in a VAE?
Instead of sampling the latent directly, you sample noise and compute z as mean plus standard deviation times noise. This makes the sampling step differentiable with respect to the encoder outputs, so gradients can flow through it during training.
Why are GANs hard to train?
The generator and discriminator are optimised against each other, so the objective is a moving target. Problems include mode collapse, where the generator covers few modes, and unstable or vanishing gradients if one network overpowers the other.
Where Generative Models fits in the Deep Learning Engineer roadmap
Generative Models belongs to the Core Architectures stage (advanced level), which takes about 4 hours. The models you must know.
Other units in this stage
Back to the full Deep Learning Engineer roadmap