Causal Inference Of Active Populations Based On Information Transfer Causal Interaction Networks

Thanks to the development of technology in recent decades,the research on various complex systems has attracted extensive attention,and one of the research directions of complex systems is to explore the behavior rules of human beings and other active groups with initiative in space movement.Understanding the mechanism of information transmission behind the collective movement of active populations can provide important insights for biology and ecology,as well as valuable models and algorithms for bionic technology and autonomous systems.However,people know little about the nature of the causal interaction network behind the information transmission of the active population,and we have almost no useful empirical information,so that it is impossible to study the causal interaction network of the information transmission once and for all simply by using a model or a combination of multiple models.Although the existing variable time delay(Lag for short)granger(VL Granger)causality inference can effectively discover the causal relationship between time series with arbitrary lag,it is still a little insufficient in the discovery of the causal relationship of the active group leadership phenomenon.This is because,when the group moves in piles,the time delay tendency between the follower and the leader will affect the choice of the best fixed lag of the whole group,until the causal relationship between the two cannot be inferred at the collective level.In this paper,we propose a Variable Relaxed-lag Granger(VRL Granger)Causality for the phenomenon of active group leadership and it can be applied in the foraging experiment of parabramis aureus.Experimental results show that compared with the existing VL Granger Causality method,the effect of VRL Granger inference method in this paper is improved by 50%.Furthermore,visual field perception model is a new solution to the active group leadership phenomenon.Variable-lag Transfer Entropy(VL Transfer Entropy)is a kind of causal inference method based on time series,compared with VL Granger causality inference method,it not only in the form of quantified system chaos change to adapt to the many organisms highly non-linear complex system requirements,and the variable delay hypothesis can subtly found in an arbitrary time delay between the time sequence affects the cause-and-effect relationship.However,VL Transfer Entropy is still insufficient in the discovery of causality of animal group information transmission,because individual animals basically rely on their visual system to obtain information from nearby individuals,and the group realizes collective behavior through such information transmission.When calculating the optimal fixed time delay,VL Transfer Entropy is directly measured by the degree of correlation of individuals in relative spatial position,which overestimates the information redundancy in the visual blind area of each individual in the fauna.In this paper,a Variable Positive-lag Transfer Entropy(VPL Transfer Entropy)based on field perception model is proposed.It can be used in the field survival of olive baboon and foraging experiment of parabramis aureus.Experimental results show that compared with VL Transfer Entropy,the effect of VPL Transfer Entropy in this paper is improved by 75% to 100%.