☰

Probability as Art

on Blog

            I went to San Francisco Museum of Modern Art (SFMOMA) yesterday, and it was my second time to visit there in this summer.

            When I was at Berkeley in 2014, SFMOMA was under construction. Unfortunately, I couldn’t get into SFMOMA, and at the beginning of 2015, I visited MOMA in New York City. After a three-year-long expansion, SFMOMA reopened on May 14, 2016, and it happened that I am in the Bay Area now.

            Among numerous impressive arts in SFMOMA, Spectrum Colors Arranged by Chance by Ellsworth Kelly stood out as a surprise for me. Since it was painted by oil on gridded wood, and Kelly arranged color for each square by totally chance to avoid his own influence on the art.

Spectrum Colors Arranged by Chance

Figure 1: Spectrum Colors Arranged by Chance


            Coincidently, this paint reminded me of a project I did last semester, Topic Modelling with Latent Dirichlet Allocation. My teammate and I implemented Latent Dirichlet Allocation, and tried to discover latent topics among documents in a corpus. At the end of our project, we visualized the result to test our implementation, and the visualization looks very similar to Kelly’s work.

Topic Modeling with Latent Dirichlet Allocation
Figure 2: Topic Modeling with Dirichlet Allocation


            Suddenly, I felt there is no boundary between art and science, maybe we can even try to discover the probability distribution of how Kelly picked the colors. Appreciation grows.