bioRxiv ScienceSearch

Biology subjects

Andres Laan

Publications and source records attributed to Andres Laan.

2 recordsLinked to original sources

Intensity matching in cuttlefish

For efficient background matching it is essential that animals closely match a set of salient visual statistics of their visual surroundings. The mean intensity of a background is a key statistic, because it can be estimated across a large range of viewing distances by a simple computation. We investigated how the dynamic neuromuscular camouflage system of the cuttlefish Sepia offcinalis responds to changes in the mean background intensity of uniform backgrounds. We find that cuttlefish adapt their body intensity in response to variations in the mean background intensity, yet show biases in their body intensity beyond what can be predicted from the limited dynamic range of their camouflage system. On sandy backgrounds of various reflectance values their uniform body patterns maintain a constant yellow hue. This color constancy may represent an example of a color prior in a colorblind animal because a yellow body color would be the optimal hue for camouflage on sands typically encountered in their natural environment. Cuttlefish adapt their appearance to the background via a dynamic process composed of a complex mixture of intensity transients spanning timescales from the subsecond to the minute range. In very young animals camouflaging on dark sands the masquerade strategy is preferred over background matching. Masquerade is implemented by combining partial background matching with frequent expression of disruptive components. We furthermore provide an objective definition of disruptive components using hierarchical clustering and automated image analysis thus highlighting the role of chromatophore activity correlations in structuring the motor output of S.officinalis.

Animal Behavior and Cognition

Echo state networks with multiple read-out modules

We propose a new readout architecture for echo state networks where multiple linear readout modules are activated at distinct time points to varying degrees by a separate controller module. The controller module, like the reservoir of the echo state network, can be initialized randomly. All linear readout modules are trained through simple linear regression, which is the only adaptive step in the modified algorithm. The resulting architecture provides modest improvements on a variety of time series processing tasks (between 5 to 50% in performance metric depending on the task studied). The novel architecture is guaranteed to perform at least as accurately as a conventional linear readout. It can be utilized as a general purpose readout method when augmentations to performance relative to the standard method is needed.\n\nPACS numbers: 05.45.Tp, 07.05.Mh

Neuroscience