Uploaded September 2023 | Updated November 2026, 3 hours ago
What happens when we treat a time counter as a sequence of feature vectors in 4D? Here, we use dimensionality reduction (t-SNE) to bring this set of points into 3D space, and the results are mind boggling..
NOTE: I only plot one point every 5 seconds, so each 1-hour sheet contains 60 x 12 points, rather than 60 * 60.
00:00 - Digital time counter as data
01:41 - Including days as a 4th variable
01:56 - t-SNE
04:00 - The initial embedding (1-second sheets)
05:18 - How we can normalize the variables to get better embeddings
08:03 - Summary and closing remarks
Github Repo: github.com/M-Lin-DM/Time-counter-embedding
This code embeds the set of vectors using t-SNE and creates animated plots of the result.
What happens when we treat a time counter as a sequence of feature vectors in 4D? Here, we use dimensionality reduction (t-SNE) to bring this set of points into 3D space, and the results are mind boggling..
NOTE: I only plot one point every 5 seconds, so each 1-hour sheet contains 60 x 12 points, rather than 60 * 60.
00:00 - Digital time counter as data
01:41 - Including days as a 4th variable
01:56 - t-SNE
04:00 - The initial embedding (1-second sheets)
05:18 - How we can normalize the variables to get better embeddings
08:03 - Summary and closing remarks
Github Repo: github.com/M-Lin-DM/Time-counter-embedding
This code embeds the set of vectors using t-SNE and creates animated plots of the result.










