Research map / Architectures and efficiency / RNNs and normalization

Self-Similarity Priors: Neural Collages as Differentiable Fractal Representations

NeurIPS 2022 · Architectures and efficiency › RNNs and normalization · approach: Learned embeddings for retrieval · 1 citations (OpenAlex)

Read the paper Check your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changed	Where	Measured	Effect
Neural Collage compressor instead of fractal compression without augmentation, low bpp	Neural Collage · DOTA held-out, low bpp	PSNR: 31.06 -> 31.30	Helped
Neural Collage instead of fractal compression without augmentation, medium bpp	Neural Collage · DOTA held-out, medium bpp	PSNR: 31.68 -> 32.12	Helped
Neural Collage instead of augmented fractal compression, medium bpp	Neural Collage · DOTA held-out, medium bpp	PSNR: 30.80 -> 32.12	Helped
Neural Collage instead of implicit COIN, low bpp	Neural Collage · DOTA held-out, low bpp	PSNR: 25.74 -> 31.30	Helped
Neural Collage instead of implicit COIN, medium bpp	Neural Collage · DOTA held-out, medium bpp	PSNR: 27.34 -> 32.12	Helped
Neural Collage compared with spectral block-DCT, medium bpp	Neural Collage · DOTA held-out, medium bpp	PSNR: 34.65 -> 32.12	Hurt
Neural Collage compared with spectral block-DCT, low bpp	Neural Collage · DOTA held-out, low bpp	PSNR: 33.22 -> 31.30	Hurt
Add rotation and color-flip augmentation of domain cells to fractal compression, low bpp	Fractal compression · DOTA held-out, low bpp	PSNR: 31.06 -> 30.51	Hurt
Add augmentation to fractal compression, medium bpp	Fractal compression · DOTA held-out, medium bpp	PSNR: 31.68 -> 30.80	Hurt
Amortize fractal encoding with a hypernetwork, test-time encoding	Neural Collage · DOTA 10 held-out images	encoding time: 100x faster	Helped
Amortize fractal encoding, including training time	Neural Collage · DOTA 10 held-out images	encoding time: up to 10x speedup	Helped
Collage VAE instead of VDVAE, rate-distortion tradeoff	Collage VAE · dynamically binarized MNIST	rate-distortion	Mixed

Closest papers by meaning

Latent Functional Maps: a spectral framework for representation alignment (NeurIPS 2024 · similarity 0.84)
Intrinsic Dimension Correlation: uncovering nonlinear connections in multimodal representations (ICLR 2025 · similarity 0.84)
How Diffusion Models Learn to Factorize and Compose (NeurIPS 2024 · similarity 0.84)
Geometry-Aware Similarity Metrics for Neural Representations on Riemannian and Statistical Manifolds (NeurIPS 2026 · similarity 0.83)
Data-free Neural Representation Compression with Riemannian Neural Dynamics (ICML 2024 · similarity 0.83)
Metric Space Magnitude for Evaluating the Diversity of Latent Representations (NeurIPS 2024 · similarity 0.83)

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft