Code repository for the paper Bayesian Experimental Design via Contrastive Diffusions

Structure of the repository:

• diffuse/: contains the source code of the Diffuse tool with the following files:

  • mixtures.py: contains the implementation of the mixture models for the test of toy diffusion
  • sde.py: Implementation of the Lin stochastic differential equations and its reverse-time counterpart used for sampling
  • images.py: tools for image processing and masking for MNIST
  • score_matching.py: implementation of the score matching loss used to train Diffusion Model
  • unet.py: implementation of the U-Net architecture used for the Diffusion Model
  • mnist_train.py: script to train the Diffusion Model on MNIST
  • conditional.py: implementation of the conditional sampling procedure
  • inference.py: utilities for conditional sampling procedure
  • optimizer.py: joint sampling and optimization procedure
  • plotting.py: utilities for plotting the results

• examples/: contains the examples of use of the Diffuse tool with a test on a toy example with Gaussian Mixtures

  • design_mnist.py: script to run the design optimization on handwritten digits retrieval
  • mixture_evolution.py: plot evolution of the mixture models with noising and denoising process

• test/: contains the tests for the Diffuse tool with a test on a toy example with Gaussian Mixtures

Visualization of the design optimization procedure:

Figure: Image reconstruction. First 6 experiments (rows): image ground truth, measurement at experiment $k$, samples from current prior $p(\theta|\mathcal{D}_{k-1}) $, with best (upper) and worst (lower) weights in each sub-row. The samples incorporate past measurement information as the procedure advances.

Comparison with random measurements:

Figure: Optimized vs. random designs: measured outcome $y$ (2nd vs. 3rd column) and parameter $\theta$ estimates (reconstruction) with highest weights (upper vs. lower sub-row).

For tests and plots of the diffusion on mixture of Gaussians:

pytest --plot