PULSE: Self-Supervised Photo Upsampling via Latent Space Exploration of Generative Models. The paper "Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling" and its source code ...Benefits of Overparameterization in Single-Layer Latent Variable Generative Models: 182: ... via Self-Supervised Learning ... space for sample efficient exploration:
Upsamples each channel of the input to a higher rate by inserting L-1 zeros between samples. Filters each channel of the upsampled data using a direct-form FIR filter. The block uses a polyphase filter implementation because it is more efficient than straightforward upsample-then-filter algorithms.