search

UCD4IDS / ContinuousWavelets.jl

wide array of continuous wavelet transforms using Julia
https://dsweber2.github.io/ContinuousWavelets.jl/dev/
Other
21 stars 7 forks source link

Morse Wavelet #20

Closed wchak closed 3 years ago

codecov-commenter commented 3 years ago

Codecov Report

Merging #20 (fbb707e) into master (2c84813) will increase coverage by 20.29%. The diff coverage is 63.93%.

Impacted file tree graph

@@             Coverage Diff             @@
##           master      #20       +/-   ##
===========================================
+ Coverage   52.92%   73.21%   +20.29%     
===========================================
  Files           7        7               
  Lines         565      631       +66     
===========================================
+ Hits          299      462      +163     
+ Misses        266      169       -97
Impacted Files Coverage Δ
src/ContinuousWavelets.jl 100.00% <ø> (ø)
src/utils.jl 65.46% <31.57%> (+16.03%) :arrow_up:
src/waveletTypes.jl 41.17% <42.85%> (-10.55%) :arrow_down:
src/createWavelets.jl 84.28% <83.33%> (+5.19%) :arrow_up:
src/CWTConstruction.jl 80.64% <100.00%> (-1.82%) :arrow_down:
... and 2 more

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 61184a9...fbb707e. Read the comment docs.

wchak commented 3 years ago

https://github.com/wchak/ContinuousWavelets.jl/blob/master/src/utils.jl#L52

I am not sure whether getMinScaling = 6 is too large, but it passes the test max([norm(nonSupported[:,i],Inf) / norm(supported[:,i],Inf) for i = 2:size(spaceWaves,2)]...) ≤ 1e-2

dsweber2 commented 3 years ago

It's passing that test because it's only constructing the father wavelet (which looks to be the test that's currently failing); there are at most 7 octaves to work with for data of size 128, so if the minimum is six, there's something odd going on with the mother wavelet construction. I would look at the space version of the unscaled mother wavelet (s=1). I suspect that its way too big, and probably something to do with the fact that they're using radian frequency instead of just frequency.