Research map / Image generation / GANs
Enabling Dark Energy Science with Deep Generative Models of Galaxy Images
Siamak Ravanbakhsh, Francois Lanusse, Rachel Mandelbaum, Jeff Schneider, Barnabas Poczos
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Use a conditional VAE to generate galaxy images conditioned on size, magnitude, and redshift | COSMOS galaxy images · COSMOS validation set | size | Improved |
| Use a conditional VAE to generate galaxy images conditioned on size, magnitude, and redshift | COSMOS galaxy images · COSMOS validation set | brightness | Improved |
| Use a conditional VAE to generate galaxy images conditioned on size, magnitude, and redshift | COSMOS galaxy images · COSMOS validation set | ellipticity distribution | Mixed |
| Add a cross-correlation penalty to decorrelate latent codes from conditions (ablation) | C-VAE · COSMOS | negative log-likelihood | Hurt |
| Use a condition-independent prior without a cross-correlation penalty (ablation) | C-VAE · COSMOS | negative log-likelihood | Hurt |
| Use a condition-dependent prior p(z|y) instead of a condition-independent prior p(z) (ablation) | C-VAE · COSMOS | negative log-likelihood | Improved |
| Use a conventional conditional GAN objective with continuous conditional variables | Conditional GAN · COSMOS | trainability | Hurt |
| Use the proposed predictor-based adversarial objective | Conditional GAN · COSMOS | condition prediction error | Improved |
| Remove the min(0, .) operation from the predictor loss (ablation) | Predictor-based conditional GAN · COSMOS | condition prediction error | Hurt |
| Use mini-batch discrimination to discourage generator collapse | Predictor-based conditional GAN · Galaxy Zoo | mode collapse | Improved |
| Add calibrated observational noise to generated C-VAE images | C-VAE · COSMOS validation set | image-statistics comparability | Improved |
| Generate images with the conditional GAN instead of the conditional VAE | Conditional generative models · COSMOS validation set | consistency of generated results | Hurt |
Closest papers by meaning
- Learning Conditional Generative Models for Phase Retrieval (JMLR 2023 · similarity 0.85)
- Unified Framework for Diffusion Generative Models in SO(3): Applications in Computer Vision and Astrophysics (AAAI 2024 · similarity 0.84)
- CosmoBench: A Multiscale, Multiview, Multitask Cosmology Benchmark for Geometric Deep Learning (NeurIPS 2025 · similarity 0.83)
- A Likelihood Based Approach to Distribution Regression Using Conditional Deep Generative Models (ICML 2025 · similarity 0.82)
- Revealing the 3D Cosmic Web through Gravitationally Constrained Neural Fields (ICLR 2025 · similarity 0.81)
- Discovering Latent Graphs with GFlowNets for Diverse Conditional Image Generation (NeurIPS 2025 · similarity 0.81)
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.
