DOC fix link to images in KMeans tutorial #24793
Conversation
|
Looking at the first figure in https://output.circle-artifacts.com/output/job/78625850-7089-4258-a7c1-53e974fd919e/artifacts/0/doc/tutorial/statistical_inference/unsupervised_learning.html:
I can't really work out how the "bad init" (bottom left) and "good" (top right) example are different. Certainly not why one is a case of bad initialisation and the other isn't. Both a very similar to each other and have a few mistakes compared to the ground truth. Is there a way to make it more obvious to the non-expert (like me) reader? |
| ground-truth (bottom right figure) and different clustering. We do not | ||
| recover the expected labels, either because the number of cluster was | ||
| chosen to be to large (top left figure) or suffer from a bad initialization | ||
| (top right figure). |
There was a problem hiding this comment.
"bottom left" and "top right" need swapping to match the labeling in the figure.
|
|
||
| .. figure:: /auto_examples/cluster/images/sphx_glr_plot_face_compress_001.png | ||
| :target: ../../auto_examples/cluster/plot_face_compress.html | ||
|
|
||
| **Raw image** |
There was a problem hiding this comment.
This headline now gets "sucked into" the code example above. My rST foo is not strong enough to suggest a fix :(
I went back in the past to understand what we tried to show: https://scikit-learn.org/0.15/tutorial/statistical_inference/unsupervised_learning.html I think the idea was to show that What I would like is to have a working example at first, as we had in the past. Then, we can open an issue to improve the current example to trigger a "real" bad init. Not sure that iris would lead to such local minimum indeed :) |
|
The text in |
|
I can do it here. It is a regression that I probably included in the last PR. |
|
I fixed the description in plot_iris. Let me know what you think. |
Co-authored-by: Tim Head <[email protected]>
closes #24791
Adapt the tutorial on unsupervised learning based on recent changes in the examples from the gallery. There is also an additional fix on one of the examples where the titles of the subfigures are not the right ones.