Conditionally Strongly Log-Concave Generative Models

Authors: Florentin Guth, Etienne Lempereur, Joan Bruna, Stéphane Mallat

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical results are shown for physical fields such as the φ4 model and weak lensing convergence maps with higher resolution than in previous works.
Researcher Affiliation Academia 1D epartement d informatique, Ecole Normale Sup erieure, Paris, France 2Courant Institute of Mathematical Sciences and Center for Data Science, New York University, USA 3Coll ege de France, Paris, France, and Flatiron Institute, New York, USA.
Pseudocode Yes Algorithm 1 Score matching for exponential families with CSLC distributions. Algorithm 2 MALA sampling from CSLC distributions.
Open Source Code Yes The code to reproduce our numerical experiments is available at https://github.com/Elempereur/WCRG.
Open Datasets Yes We used down-sampled versions of the simulated convergence maps from the Columbia Lensing Group (http://columbialensing.org/; Zorrilla Matilla et al., 2016; Gupta et al., 2018) as training data.
Dataset Splits No The paper mentions generating images of specific sizes and using them as training data, but it does not provide explicit train/validation/test splits, percentages, or sample counts.
Hardware Specification No The paper mentions that
Software Dependencies No The paper mentions "Py Torch" and "Py Wavelets" but does not provide specific version numbers for these or other software dependencies.
Experiment Setup Yes The MALA step sizes δj are adjusted to obtain an optimal acceptance rate of 0.57. Depending on the scale j, the stationary distribution is reached in Tj 20 400 iterations from a white noise initialization. We used a qualitative stopping criterion according to the quality of the matching of the histograms and power spectrum.