You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I was trying to help someone use alphaCSC and realized that it's not straightforward to understand which class to use when. We are also missing some guidelines how to tweak certain parameters (e.g., which solver to choose?).
Empirically it seems that LGCD is better but it is not possible to use it from learn_d_z. I think this function should be deprecated and either GreedyCDL or BatchCDL should be supported also for univariate signals. Zen of Python says:
There should be one-- and preferably only one --obvious way to do it.
The GreedyCDL class should be documented in api.rst. And to use it for univariate signals, one needs to set rank1=False and uv_constraint='joint'.
In the same vein, split_signal is supported only for 2D signals at the moment, but extending it to 1D data will increase the user base with very little effort. Some efforts in making these things consistent would be great!
The text was updated successfully, but these errors were encountered:
I was trying to help someone use
alphaCSC
and realized that it's not straightforward to understand which class to use when. We are also missing some guidelines how to tweak certain parameters (e.g., which solver to choose?).Empirically it seems that LGCD is better but it is not possible to use it from
learn_d_z
. I think this function should be deprecated and eitherGreedyCDL
orBatchCDL
should be supported also for univariate signals. Zen of Python says:The
GreedyCDL
class should be documented in api.rst. And to use it for univariate signals, one needs to setrank1=False
anduv_constraint='joint'
.In the same vein,
split_signal
is supported only for 2D signals at the moment, but extending it to 1D data will increase the user base with very little effort. Some efforts in making these things consistent would be great!The text was updated successfully, but these errors were encountered: