Conversation
|
@naveedishaque what do you think regarding the structure? |
| @@ -0,0 +1,6 @@ | |||
| channels: | |||
| - conda-forge | |||
| - defaults | |||
There was a problem hiding this comment.
I would remove defaults. Shouldn't be needed here
| @@ -0,0 +1,6 @@ | |||
| channels: | |||
| - conda-forge | |||
| - defaults | |||
There was a problem hiding this comment.
I would remove defaults. Shouldn't be needed here
| } | ||
|
|
||
| # Randomize labels | ||
| df_randomized <- data.frame(label = sample(df$label)) |
There was a problem hiding this comment.
We might have an additional colum, in this dataframe that splits the label into high and low confidence. Should that be shuffled too?
@naveedishaque
There was a problem hiding this comment.
I think you mean "if we shuffle the label we should also shuffle the confidence"?
My feeling is to keep a low confidence spot as a low confidence spot even after label shuffling.
Does that make sense?
There was a problem hiding this comment.
Yeah, good point. In that case the code still needs to be adjusted to keep the additional columns untouched. Make sure only the labels are shuffled and the rownames still match all the other existing columns
There was a problem hiding this comment.
Sounds good. I'll add the changes :)
They should be separate |
|
We need to figure out how to make use of it i.e. how tpo connect this to the metrics and trace back that it is a simulation. I think this probably requires a bit more thought on the workflow side. Also wondering if we need a separate folder for simulations @naveedishaque |
3 participants
Implementation of shuffling labels (closes issue #80) and shuffling coordinates (closes issue #79).
Notes