Skip to main content

t-SNE

t-SNE & UMAP ... They're just art—Lior Pachter

An exploration of t-distributed stochastic neighbor embedding (t-SNE) as generative mechanism. t-SNE is a widely used nonlinear dimension reduction technique. Here, however, it is used entirely for artistic purposes. The input data consists of arbitrarily generated groups of points. The groups retain their identity after t-SNE has been run, and group identity is used to color individual points in the final piece. Thus, the extent to which t-SNE retains or destroys the original group identity shapes the visual appearance of each individual piece.

The entire series was generated by a single R program, using 99 consecutive random seeds. The respective random seed is indicated in the name of each piece.

attach_money
keyboard_arrow_down
to
search
search
search
keyboard_arrow_down
keyboard_arrow_down

99 items

t-SNE #5
Price
0.1
lock

0

t-SNE #6
Price
0.1
lock

0

t-SNE #11
Price
0.1
lock

0

t-SNE #19
Price
0.1
lock

0

t-SNE #14
Price
0.1
lock

0

t-SNE #15
Price
0.1
lock

0

t-SNE #17
Price
0.1
lock

0

t-SNE #18
Price
0.1
lock

0

t-SNE #8
Price
0.1
lock

0

t-SNE #10
Price
0.1
lock

0

t-SNE #2
Price
0.1
lock

0

t-SNE #3
Price
0.1
lock

0

t-SNE #4
Price
0.1
lock

0

t-SNE #25
Price
0.1
lock

0

t-SNE #13
Price
0.1
lock

0

t-SNE #12
Price
0.1
lock

0

t-SNE #21
Price
0.1
lock

0

t-SNE #23
Price
0.1
lock

0

t-SNE #24
Price
0.1
lock

0

t-SNE #16
Price
0.1
lock

0

t-SNE #22
Price
0.1
lock

0

t-SNE #9
Price
0.1
lock

0

t-SNE #1
Price
0.1
lock

0

t-SNE #7
Price
0.1
lock

0

t-SNE #30
Price
0.1
lock

0

t-SNE #29
Price
0.1
lock

0

t-SNE #28
Price
0.1
lock

0

t-SNE #35
Price
0.1
lock

0

t-SNE #34
Price
0.1
lock

0

t-SNE #33
Price
0.1
lock

0

t-SNE #38
Price
0.1
lock

0

t-SNE #43
Price
0.1
lock

0

t-SNE

t-SNE & UMAP ... They're just art—Lior Pachter

An exploration of t-distributed stochastic neighbor embedding (t-SNE) as generative mechanism. t-SNE is a widely used nonlinear dimension reduction technique. Here, however, it is used entirely for artistic purposes. The input data consists of arbitrarily generated groups of points. The groups retain their identity after t-SNE has been run, and group identity is used to color individual points in the final piece. Thus, the extent to which t-SNE retains or destroys the original group identity shapes the visual appearance of each individual piece.

The entire series was generated by a single R program, using 99 consecutive random seeds. The respective random seed is indicated in the name of each piece.

attach_money
keyboard_arrow_down
to
search
search
search
keyboard_arrow_down
keyboard_arrow_down

99 items

t-SNE #5
Price
0.1

0

t-SNE #6
Price
0.1

0