Mapping the latent space of neural media (2018)

SK. Choi, Postcards from an ‘other’, 2018, parameter range style-transfer digital composite image.

“Neural media” emerges from deep learning image classification technology. These technologies, broadly construed, group information based on multi-dimensional statistical sampling to build high-level pattern descriptions of diverse input data. Defining information as recombinant patterns of levels of texture-structure enables that same information to be remixed or reassembled in ways that the embodied being would not otherwise encounter. These spaces of ambiguity in the data field are called “latent” spaces, for the embedded potentiality inherent in the discrete interpretation central to computational modelling. Exploring latent spaces of computational potentiality I map parameter against parameter (here, content-weight against style-scale, or the relative appearance of information statistically defined as form-related and other information statistically defined as color-distribution), leaving affective markers as I venture into unexplored regions where there are rumored to be dragons still. One treads carefully in the algorithmic landscape, and it is easy to lose one’s way. Like Borges’ (1948/2007) garden of forking paths the virtual leads by ambiguous circuits back to places that are strangely familiar but never seen before. I created many of these studies to make initial determinations of the personally resonant parameter space of the style-transfer algorithm. See for better resolution.

SK. Choi, Incrementation #1, Green phase, 2018, style-transfer digital image.
SK. Choi, Incrementation #2, Blue phase, 2018, style-transfer digital image.

In these images I enter the space of an aesthetics of optimization. Color distribution in the composition is scaled and cross-correlated with the content statistics of my source images which are weighted for degree of influence. Grids of parameter incrementation across a range of such image scales produce surfaces promoting reflective pareidolic phenomena. The same space is perceived through different depth clues, suggesting movement.

Examples of typical “content” (left) and “style” (right) images used in Incrementation #1 and #2. The right-hand image is a composite of three (red, yellow, blue) versions of the style images applied, hence the different color casts. Various electronic components were used for content imagery (© SK. Choi 2018).

These two images are typical of my juxtaposition of content images of industrial control with the structuring disorder of embodied experience that models style.

Borges, J. L. (1948/2007). Labyrinths: Selected Stories & Other Writings. D. A. Yates and J. E. Irby (Eds.). New York: New Directions.

About Suk Kyoung Choi

artist / researcher

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: