Open Juphex opened 1 month ago
The tick plots tend to give people the sense that the axes and scales are quite meaningful, when in general they aren't -- it is relative positions that matter. For that reason I prefer to suppress them as it generally leads people to better interpretations.
On Tue, Jul 23, 2024 at 8:02 PM Christian Salamut @.***> wrote:
Why do you disable the axis ticks of the plot ? See https://github.com/lmcinnes/umap/blob/126a31f31ff78c5542d479335a9f51cf38d68c6c/umap/plot.py#L700
Isn't it quite useful to have it to be able to compare between plots because the plot is not normed (e.g. between 0 and 1)? I disabled it locally and I think it would be nice to have it as a feature, right?
Best
— Reply to this email directly, view it on GitHub https://github.com/lmcinnes/umap/issues/1141, or unsubscribe https://github.com/notifications/unsubscribe-auth/AC3IUBITQCR33V5EH3R5RIDZNYSZ5AVCNFSM6AAAAABLKCPT6CVHI2DSMVQWIX3LMV43ASLTON2WKOZSGQZDINZYGQ2DEMQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>
Thank you!
When comparing similar neural network models with minor differences, given the same inputs and visualizing features at the same point in the network, should we enable the axis and use the same axis range for both plots to facilitate comparison?
If you want to make direct comparisons then yes -- umap.plot.points does just output a matplotlib object so you can take the result and apply the usual matplotlib operations as you wish to re-add back in ticks etc. as desired.
Why do you disable the axis ticks of the plot ? See https://github.com/lmcinnes/umap/blob/126a31f31ff78c5542d479335a9f51cf38d68c6c/umap/plot.py#L700
Isn't it quite useful to have it to be able to compare between plots because the plot is not normed (e.g. between 0 and 1)? I think it would be nice to have it as a feature, right?
Best