Conversation
- coadd_maps_pixels now also returns the per-pixel ILC weights, and
coadd(..., return_weights=True) returns them alongside the map.
- needlet_coadd writes the weights estimated for 'coadd' to
{out_root}wavelet_weights_scale_{k}_{tag}.fits (removed with
delete_intermediate like the covariance maps).
- New 'cmb_weights' nmap label applies the weights to ones maps and
returns the per-scale weight sums (no wave2map) in
outmaps['cmb_weights_coadd'].
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Runs a block-smoothed CMB NILC on the synthetic test_pipeline dataset with nmap_labels=["cmb_weights"] and prints, per wavelet scale, the range of cmb_weights_coadd and the sum over tags of the written weight maps (both should be 1 wherever any map contributes). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
needlet_coadd no longer accepts 'cmb_weights' in nmap_labels. Per-scale ILC weight maps are still written to disk; with check_cmb_weights=True, each scale's weights are checked to sum to 1 in every covered pixel (CMB solution), raising ValueError otherwise. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
For gaussian and tophat covariance smoothing, the smoothed products and local means averaged in zeros from outside each map's footprint, so the covariance was biased low (by up to ~50%) near mask edges. Normalizing each pair by its own smoothed overlap fixes the bias but makes the per-pixel covariance indefinite when more than two footprints overlap. Instead, at each pixel every entry is now estimated over the common mask M_T of the set T of maps covering that pixel, C_ij = S[M_T d_i d_j] / S[M_T], which is unbiased at edges and stays positive semi-definite for a non-negative kernel. The local mean used for smooth_mean_cov is normalized by each map's own mask. Block smoothing is unchanged. Add tests/test_cov_edges.py, which checks that the covariance is unbiased and positive definite for three maps with overlapping edges. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The coadder returns zero wherever no mask covers a pixel, so the taper region and source holes of the base map were zeroed in each wavelet scale. The sharp cut leaked into the footprint on reconstruction, giving a ~6% rms signal error at low ell even far from the mask edge. Those pixels now take the base map's own wavelet coefficients; the final base mask is still applied afterwards. Add tests/check_two_map_subset.py, a two-map test (wide noisy base map plus a deeper map on a subset footprint) that checks per-scale ILC weights against the analytic inverse-noise-variance solution, signal preservation and noise of the coadd for block, gaussian and tophat covariance smoothing. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Changes include the following (most of heavy lifting done by Claude Opus 5.5):