A Deep Learning Reconstruction of Animal Behaviour

The development of accelerometers and GPSs has enabled to instrument individual birds, in particular seabirds, and monitor the behavior of birds at sea in unprecedented detail. This is providing vast amounts of temporal series data, with the consequent challenge to develop methods appropriate for its analysis in terms that can be scientifically interpretable.

On the other hand, recent developments of specific types of Deep Learning Networks for temporal series analysis, Transformer Networks or Convolutional Networks have provided remarkable results in identifying patterns in vast datasets in a wire variety of applications.

The project we propose here is precisely aimed at developing advanced tools to identify patterns of behavior based on Deep Learning networks, overcoming a number of known problems specific to the analysis of this kind of data: vast amounts of data, vast variability specifics as a function of the kind of behavior, in most cases lack of external validation for individual instances of behaviour.

This is a collaboration between two research groups. The major research lines and publications records of each group are outlined in the following links:

www.gonzalez-solis.com