Modelling hysteresis in GeNie

[Richard Dean]

Can anyone advise me how I might represent a hysteretic process in GeNie?

For example, the probability that a lake will be eutrophic depends, in large part, upon its phosphorus loading (and resultant water column concentration). However, because of system feedbacks the probability is different depending upon whether the P loading is increasing from low to high level or declining from high to low level. The relationship between P loading and eutrophication risk is hysteretic. So given a particular P loading, the probability that a lake is eutrophic could take two different values depending upon whether the "forwards" or "reverse" process is being considered. How can a Bayesian network model this?

Thanks!

[marek [BayesFusion]]

Can anyone advise me how I might represent a hysteretic process in GeNie?

For example, the probability that a lake will be eutrophic depends, in large part, upon its phosphorus loading (and resultant water column concentration). However, because of system feedbacks the probability is different depending upon whether the P loading is increasing from low to high level or declining from high to low level. The relationship between P loading and eutrophication risk is hysteretic. So given a particular P loading, the probability that a lake is eutrophic could take two different values depending upon whether the "forwards" or "reverse" process is being considered. How can a Bayesian network model this?

Have you looked at the DBN modeling in GeNIe? Start with Network/Enable Temporal Plate. You can draw arcs between nodes with time delay and feedback loops. I suspect that this way you can do what you want. Please let us know if this answer is off mark.
Cheers,

Marek

[Richard Dean]

Thanks Marek, to be honest I haven't progressed to DBN modelling yet. However, I've given the problem more thought. Maybe a simple approach is just to add a decision node with two states: forward process and reverse process. I'll give it a try.