Closed aolney closed 4 years ago
I wonder about KNN in this one. Thoughts so far:
KNN has a probabilistic interpretation, but can be conceptualized/used without probability (majority vote)
KNN is much simpler than LDA, so perhaps should come before it in the notebook
It may make sense to pull LDA into its own notebook
1) I would use the non probability view. I tried to refrain from dealing with probabilities in the programming implementation. I believe these concepts are too strong for this level.
2) I think LDA needs to be in a notebook with PCA.
Tasha
From: Andrew M Olney notifications@github.com Sent: Wednesday, June 10, 2020 9:36:04 AM To: memphis-iis/datawhys-content-notebooks datawhys-content-notebooks@noreply.github.com Cc: Natasha Sahr nsahrsimonis@gmail.com; Assign assign@noreply.github.com Subject: Re: [memphis-iis/datawhys-content-notebooks] Notebook: KNN classification (#6)
I wonder about KNN in this one. Thoughts so far:
KNN has a probabilistic interpretation, but can be conceptualized/used without probability (majority vote)
KNN is much simpler than LDA, so perhaps should come before it in the notebook
It may make sense to pull LDA into its own notebook
— You are receiving this because you were assigned. Reply to this email directly, view it on GitHubhttps://github.com/memphis-iis/datawhys-content-notebooks/issues/6#issuecomment-642014022, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AHEKYFSPDXV4PFZ2KBHB4W3RV6D4JANCNFSM4M2B3DFA.
Another thought; don't have time to fix before 10am: the two confusion matrix graphics use different axes. One has actual class on the X and the other has it on the Y. This may create confusion.
I will fix but won’t be before 10am. My apologies for that.
From: Andrew M Olney notifications@github.com Sent: Wednesday, June 10, 2020 10:04:57 AM To: memphis-iis/datawhys-content-notebooks datawhys-content-notebooks@noreply.github.com Cc: Natasha Sahr nsahrsimonis@gmail.com; Assign assign@noreply.github.com Subject: Re: [memphis-iis/datawhys-content-notebooks] Notebook: KNN classification (#6)
Another thought; don't have time to fix before 10am: the two confusion matrix graphics use different axes. One has actual class on the X and the other has it on the Y. This may create confusion.
— You are receiving this because you were assigned. Reply to this email directly, view it on GitHubhttps://github.com/memphis-iis/datawhys-content-notebooks/issues/6#issuecomment-642032736, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AHEKYFXAWXZAXIHYBVB4OL3RV6HITANCNFSM4M2B3DFA.
No worries at all - this thread is for tracking so we get around to it eventually :thumbsup:
Another note on this. I cut the AUC/ROC graphs from the AM notebook, but if we pull out the LDA portion from that notebook, I think we could put them back in. I like them a lot, and they complement the didactic portion nicely. It was just too much material to keep them in for the morning session with what was already there.
HI Andrew, Yes I have worried over this notebook a lot. I like your suggestion of putting the KNN first and I am ok with making LDA a separate notebook or just leaving it out. Best, Dale
From: Andrew M Olney notifications@github.com Sent: Wednesday, June 10, 2020 8:36 AM To: memphis-iis/datawhys-content-notebooks datawhys-content-notebooks@noreply.github.com Cc: Dale D Bowman (ddbowman) ddbowman@memphis.edu; Assign assign@noreply.github.com Subject: Re: [memphis-iis/datawhys-content-notebooks] Notebook: KNN classification (#6)
I wonder about KNN in this one. Thoughts so far:
KNN has a probabilistic interpretation, but can be conceptualized/used without probability (majority vote)
KNN is much simpler than LDA, so perhaps should come before it in the notebook
It may make sense to pull LDA into its own notebook
— You are receiving this because you were assigned. Reply to this email directly, view it on GitHubhttps://github.com/memphis-iis/datawhys-content-notebooks/issues/6#issuecomment-642014022, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AOUET7RE257YK6T5KH744VTRV6D4JANCNFSM4M2B3DFA.
I agree. I think the ROC/AUC would probably be more useful, especially since 2 of the interns are planning on going to med school.
From: Andrew M Olney notifications@github.com Sent: Wednesday, June 10, 2020 11:58 AM To: memphis-iis/datawhys-content-notebooks datawhys-content-notebooks@noreply.github.com Cc: Dale D Bowman (ddbowman) ddbowman@memphis.edu; Assign assign@noreply.github.com Subject: Re: [memphis-iis/datawhys-content-notebooks] Notebook: KNN classification (#6)
Another note on this. I cut the AUC/ROC graphs from the AM notebook, but if we pull out the LDA portion from that notebook, I think we could put them back in. I like them a lot, and they complement the didactic portion nicely. It was just too much material to keep them in for the morning session with what was already there.
— You are receiving this because you were assigned. Reply to this email directly, view it on GitHubhttps://github.com/memphis-iis/datawhys-content-notebooks/issues/6#issuecomment-642136780, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AOUET7XWZRSF53YYRGEQ3ADRV63S7ANCNFSM4M2B3DFA.
New confusion matrix images added with https://github.com/memphis-iis/datawhys-content-notebooks/pull/38
Temporarily closing, but eventually we will reopen to make discussed changes
See the spreadsheet for details
Ideas/prereqs: Classification and prediction, causality and experiments, Functions, estimating functions, decision boundaries, supervised vs unsupervised, features, error, error minimization,
Direct link https://jupyter.olney.ai/hub/user-redirect/git-pull?repo=https%3A%2F%2Fgithub.com%2Fmemphis-iis%2Fdatawhys-content-notebooks&subPath=KNN-classification.ipynb&app=lab