Interactive Evolution of Swarms for the Visualisation of Consumptions

Information Visualisation studies how visual representations can help in the understanding of hidden patterns in large amounts of data. In the Data Aesthetics field, data is used to create visualizations which are more concerned with aesthetics. In this work, we developed a framework to explore the aesthetic dimension of a functional visualisation model. In concrete, we will explore a dataset with the consumption of the Portuguese people in one of the biggest retail companies in Portugal. Our main goal is to use Interactive Evolutionary Computation (IEC) to evolve the parameterisation of a visualisation model, enabling the user to explore new possibilities and to create different aesthetics over the consumption patterns. The developed system is able to create a wide diversity of emergent visual artefacts that can be intriguing and aesthetically appealing for the user (Fig. 1).

 
 

Figure 1

A small set of the variety of visual artefacts that the evolutionary system can create. Depending on the user guidance this set can be extended to other visual solutions not presented in this image.


 
 

Evolutionary Approach

 

To improve the swarm system detailed in [1], we apply an Evolutionary Algorithm (EA) to increase the degrees of freedom in the creation of the visual models, enabling a diverse range of solutions. Our intention with this new approach is to develop visual artefacts that are continuously readjusting to the intentions of the user and to enable the user to explore artefacts not imagined by him/her. In this way, the company can deploy a system that will evolve and adapt to different audiences’ aesthetic preferences.

 

Evolutionary Algorithm

 

The visualisation model described in [1] is easy to understand and use, yet it requires the definition of several parameters to create visual artefacts that can be appealing for the user. To aid in this task, we propose a framework based on EA. To evolve the swarm system, we will be searching for the best combination of the following system’s parameters: (i) the separation, alignment, and cohesion forces; (ii) the minimum and maximum radius; (iii) the use of a global or local normalisation; and (iv) the representation modes (lines, circles, transparency, sorted circles). Note that the boids’ size is always mapped according to the consumption value on the data. In the following subsections, we present the parameters used to evolve the visual artefacts and the used genetic operators.

 

Usage Scenarios

 

Through IEC, we explored the evolutionary system based on the user’s preferences. We implemented a simple interface to enable the interaction with the system. These explorations were based on three different objectives. In the first exploration, the goal is to evolve solutions with specific parameterisation attributes: the boids must be represented with filled circles, use the local normalisation, and have a zigzagging pattern (Fig. 2).
 

Figure 2

Chosen individuals by the user during the interactive evolution. In this example, the user aimed to create artefacts that use circles and have a balanced colour palette. The top individual is the final choice past the 10 generations. In the row below, the first artefact is the first choice in the first generation, the individual in the middle was gathered in the third generation, and the rightmost, the chosen one in the sixth generation.


 
In the second, the goal is to attain solutions that must intersect the functional dimension, enabling the readability of the artefacts (Fig. 3). Finally, for the third exploration, there is no predefined objective, so the solutions must be diversified according to the user’s taste (Fig. 4).
 
Figure 3

Chosen individuals by the user during the interactive evolution. In this example, the user aimed to create artefacts that can be placed in the functional dimension. The top individual is the final choice after 10 generations. In the row below, the first artefact is the first choice in the first generation, the individual in the middle was gathered in the third generation, and the rightmost, the chosen one in the sixth generation.


 
For the first two explorations, the user must have some experience on how the system works, its parameters and the data. In the last, the user can have no experience with the data nor with the parameterisations, the user only explores the possibilities and creates artefacts suitable for his/her own taste. As the generations evolve, the user chooses only the visual artefacts that visually intrigue, amaze, and/or correspond with his/her preferences.
 
Figure 4

Chosen individuals by the user during the interactive evolution. In this example, the user has no predefined target solution, choosing only the solutions which fit his/her tastes. The top individual is the final choice past the 10 generations. In the row below, the first artefact is the first choice in the first generation, the individual in the middle was gathered in the third generation, and the rightmost, the chosen one in the sixth generation.

 

References

 

[1] Maçãs, C., Cruz, P., Martins, P., Machado, P.: Swarm systems in the visualization of consumption patterns. In: Yang, Q., Wooldridge, M. (eds.) Proceedings of the Twenty-Fourth International Joint Conference on Artificial Intelligence, IJCAI 2015, Buenos Aires, Argentina, July 25-31, 2015. p. 24662472. AAAI Press (2015)

 
 

To appear in

Catarina Maçãs, Nuno Lourenço and Penousal Machado. Interactive Evolution of Swarms for the Visualisation of Consumptions. ArtsIT 2018 – 7th EAI International Conference: ArtsIT, Interactivity & Game Creation, 2018

Author

Catarina Maçãs

Nuno Lourenço

Penousal Machado