Fast importance sampling for model robustness evaluation. Tools for DES 3x2pt extensions.
pip install git+https://github.com/des-science/fastismore- Run your favorite sampler in cosmosis with
extra_output = ... sigma_crit_inv_lens_source/sigma_crit_inv_1_1 sigma_crit_inv_lens_source/sigma_crit_inv_1_2 ... data_vector/2pt_theory#639where #639 should be replaced with the size of your data vector after scale cuts, and the sigma_crit_inv_i_j factors should include all combinations of lens bin i with source bin j.
-
Run
fastis-sampleto compute importance weights for the new data vector. -
Plot results using
fastis-plotor your favorite script.
If you prefer to work in a notebook environment, chains can be loaded as in the following example:
import fastismore
import fastismore.plot
baseline = fastismore.Chain('baseline_chain.txt')
contaminated = fastismore.ImportanceChain('importance_weights.txt', baseline)
fastismore.plot.plot_2d(param1, param2, [baseline, contaminted], truth, labels, sigma=0.3))For more use cases, check the examples directory.