Automatic analysis of rodents behaviour has received growing attention in recent years as rodents are the reference species for large scale pharmacological and genetic screenings. In this paper we propose a new method to identify prototypical high-level behavioural patterns which go beyond simple atomic actions. The method is embedded in a data mining pipeline thought to support behavioural scientists in exploratory data analysis and hypothesis formulation. A case study is presented where the method is capable of learning high-level behavioural prototypes which help discriminating between two strains of mouse having known differences in their behaviour.
|Titolo:||Dirichlet process mixtures of multinomials for data mining in mice behaviour analysis|
|Data di pubblicazione:||2014|
|Appare nelle tipologie:||04.01 Contributo in atti di convegno|