Visualizing the Fortune 500 by similarity.

Description

I am especially interested in the intersection of A.I. and design. In this experiment, I use A.I. techniques to organize and visualize the logos of the Fortune 500 companies, by visual similarity. A task that would be both time-consuming and difficult for a human takes the computer mere moments. The outcome is a live visualization, where calculations are done in the browser, enabling the viewer to change the parameters of the underlying algorithms—altering the outcome.

Thank you to everyone who helped me with this project. Special thanks to Fortune Magazine, Ryan Keisler, Hugh Dubberly, Cody Wackerman, Monica Miller, and DDO.

Categories

Computational Design

Information Design

Neural Network

Machine Learning

t-SNE

Year

2018

↑ Fig. 1: A solution develops

A timelapse of 600 t-SNE iterations. The solution generally improves with each iteration. Subject to diminishing returns.

↑ Fig. 2: Align to grid

The logos are aligned to a grid. The Linear assignment problem (LAP) algorithm is a greedy algorithm, and the solution is not guaranteed to be optimal.

asdasd

↑ Fig. 3: Final composite

Here is one of many possible outcomes. The solution is not deterministic, and will vary slightly each time the visualization is loaded.

An interactive experience

Fig 4: The introduction outlines the process, step by step. Buying some time for the calculations to finish in the background.
Fig 4: The introduction outlines the process, step by step. Buying some time for the calculations to finish in the background.

Fig 4: The introduction outlines the process, step by step. Buying some time for the calculations to finish in the background.

1 / 2
Fig 5: The visualization is interactive. Users can zoom, pan, and manipulate the solution space.
Fig 5: The visualization is interactive. Users can zoom, pan, and manipulate the solution space.

Fig 5: The visualization is interactive. Users can zoom, pan, and manipulate the solution space.

1 / 2
Fig 6: A live visualizations makes it possible to change the parameters of the underlying algorithms, altering the outcome—highlighting different patterns.
Fig 6: A live visualizations makes it possible to change the parameters of the underlying algorithms, altering the outcome—highlighting different patterns.

Fig 6: A live visualizations makes it possible to change the parameters of the underlying algorithms, altering the outcome—highlighting different patterns.

1 / 2