Weekly journal 2024-11-10~2024-11-17
prev: Weekly journal 2024-11-06~2024-11-10
I'll switch back to a short but weekend switching cycle to get back on track.
I wrote last time, "Things to do this week."
We've done a lot of broad listening from opinion vectors, so we want to do broad listening from text.
I've thought about it, but I haven't moved my hands on it.
Broad Listening from Text Format Data
However, now that we have the final data for the public opinion map, I'd like to analyze that as well.
Public Opinion Map 3970 UMAP
I had trouble dealing with the high-dimensional mud pie in the center.
Should UMAP results be clustered?
Taught: Careful clustering of tSNE results.
The idea at the end of "[Should UMAP results be clustered?" for handling high-dimensional mud pie was deemed nonsensical as a result of calling "Careful clustering of tSNE results". I'm wondering what I should do about the next action.
Solving simple problems for better understanding.
Distance of high-dimensional normal distribution has mean √D, variance 0.7
antagonistic dimension
Mathematical multidimensional spiky spikes
As for broad listening from text format data, if the clusters are not appropriate, the AI explanation of clusters will also be poor.
What to do this week
For now, let's take it easy on the UMAP for the public opinion map since it is a stretch goal, and update the WebUI with the final version of the data once it is updated ✅.
There may be more high-priority discussions at the meeting scheduled for later in the day.
It wasn't much.
If you have time, take time to think about Broad Listening from Text Format Data, External Brain, Teaching AI the KJ Method, etc. in particular.
I'm getting more referrals from all over the world, and I'm being asked to write about various things, but I'm too busy to write in English, so I should proceed with the Scrapbox Automatic Translation project, and maybe I should become fluent in Chinese, Malay, and Vietnamese as well as English. Maybe it would be better to become not only English, but also Chinese, Malay, and Vietnamese.
Related: Look East Policy.
Maybe we can improve the UI of mem.nhiro.org.
It looks like we're making a lot of progress, so let's get a grasp of the current status of Plurality's Japanese translations.
Also try [Difference between DBSCAN and HDBSCAN
---
This page is auto-translated from /nishio/週記2024-11-10~2024-11-17 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.