Unsupervised non-linear neural networks capture aspects of floral choice behaviour

Two unsupervised neural networks were tested to understand the extent to which they capture elements of bumblebees’ unlearned preferences towards flower-like visual properties. The networks, which are based on Independent Component Analysis and Feature-Extracting Bidirectional Associative Memory use images of test-patterns that are identical to ones used in behavioural studies. While both models show consistency with behavioural results, the ICA model matches behavioural results substantially better in terms of image reconstruction quality of radial and concentric patterns, and foliage background. Both models generated a novel prediction of an interaction between spatial frequency and symmetry. These results are interpreted to support the hypothesis that flower displays are adapted to pollinators’ information processing constraints. 1 Information Processing in Bumblebees Bees use visual information to discover their first rewarding flower, but it’s not clear how the visual system aids in this discovery. Hymenoptera species including bumblebees and honeybees are frequently studied to investigate informationprocessing biases of the visual system. Unlearned visual preferences by bees are usually studied by decomposing a natural flower into its visual constituents and pitting two or more visual properties against one another in choice experiments [1]. Some of the studied components include colour [2], shape [3], symmetry [4], foliage background complexity [5], and pattern positioning [6]. Unlearned floral preferences have been tested in many pollinator species, but the idea that the preferences are a by-product of the information processing properties of environmental information has only been recently suggested [7]. Nectar guides (i.e. radial, sunburst pattern) and floral symmetry are often suggested to be an adaptation, but the distinction of whether the adaptation belongs to the plant or the pollinator, or both, is not specified. Our hypothesis is that nectar guides and symmetric floral displays are an adaptation by the plant to exploit a information-processing constraints in pollinators’ nervous system. The focus of this paper is to compare empirical results with unsupervised neural networks in relation to (a) pattern shape and positioning, (b) foliage background and (c) symmetry and spatial frequency. The networks are evaluated by generating filters from experimental stimuli that are identical in appearance to those used in behavioural studies. 149 ESANN 2013 proceedings, European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning. Bruges (Belgium), 24-26 April 2013, i6doc.com publ., ISBN 978-2-87419-081-0. Available from http://www.i6doc.com/en/livre/?GCOI=28001100131010.