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
We're using your work for an upcoming release of cbioportal.org. We wanted more control over the rendering, so we used your exposed algorithms for computing layout, and manually rendering the circles using that layout. This has been a godsend, thank you.
But I discovered an issue - sometimes, in a particularly tight layout situation, the circles would layout so as to appear that an actually empty intersection is not empty.
For example:
I believe this is better because as described in the screenshot, it weights small sets equally. I don't have time to open a PR but I wanted to drop this idea in because I think it's a good one.
The result is:
The text was updated successfully, but these errors were encountered:
adamabeshouse
changed the title
Default loss function should use log ratio loss, not difference
Default loss function should use log ratio, not difference
May 15, 2019
adamabeshouse
changed the title
Default loss function should use log ratio, not difference
Default loss function should use log ratio, not subtraction
May 15, 2019
We're using your work for an upcoming release of cbioportal.org. We wanted more control over the rendering, so we used your exposed algorithms for computing layout, and manually rendering the circles using that layout. This has been a godsend, thank you.
But I discovered an issue - sometimes, in a particularly tight layout situation, the circles would layout so as to appear that an actually empty intersection is not empty.
![image](https://user-images.githubusercontent.com/636232/57811887-6a31ca80-7739-11e9-9a33-1a9b32651155.png)
For example:
I was able to fix this on our end by copying over the loss function but modifying this line
https://github.com/benfred/venn.js/blob/master/src/layout.js#L416
to turn it into
![image](https://user-images.githubusercontent.com/636232/57811685-f42d6380-7738-11e9-8fd5-645599e24c7a.png)
I believe this is better because as described in the screenshot, it weights small sets equally. I don't have time to open a PR but I wanted to drop this idea in because I think it's a good one.
The result is:
![image](https://user-images.githubusercontent.com/636232/57811903-73bb3280-7739-11e9-9b21-2fd8f4e4d15e.png)
The text was updated successfully, but these errors were encountered: