HI Marek;
i want to discretize a equation for DBN. i foud this topic viewtopic.php?t=4719 and your predator-prey models and your files.
are you used from discretezied CPD node F conditional F-1 and R-1 for define CPD node F in t>1 for DBN? and the same for R ?
i mean this table
in my case i want to discretize this equation. this equation is time-invariant. a(t) = dt
K and M are lognormal and Kt=K0 and Mt=M0. and others are deterministic; and as you can see in equation dt is depending to previouse time step too (dt_1).
is my model is true for discretize dt and use it's CPD IN DBN for t>1 in ?
thank you
discretize equation node for DBN
-
- Site Admin
- Posts: 432
- Joined: Tue Dec 11, 2007 4:24 pm
Re: discretize equation node for DBN
Hi Bahman,
You are correct -- I used the CPT created by automatic discretization for copying and pasting into the discrete nodes into the DBN.
You can do the same for any dynamic model -- you just need to reflect on the correct structure, i.e., what influences what. You can get this information from the structure of the differential/difference equations in the dynamic model. I had the Lotka-Volterra equations to start with. I assume you have also a system of differential equations that you can start with?
The equation that you reproduced is not time-invariant and is a difference equation. Very clearly a_t depends on a_{t-1}. Please reflect on which of the variables/symbols in the equation are constants/parameters and which are actually dynamic variables. Then how they interact. This is the best I can do for you. I will be happy to get hired to do this work for you :-).
I hope this helps,
Marek
You are correct -- I used the CPT created by automatic discretization for copying and pasting into the discrete nodes into the DBN.
You can do the same for any dynamic model -- you just need to reflect on the correct structure, i.e., what influences what. You can get this information from the structure of the differential/difference equations in the dynamic model. I had the Lotka-Volterra equations to start with. I assume you have also a system of differential equations that you can start with?
The equation that you reproduced is not time-invariant and is a difference equation. Very clearly a_t depends on a_{t-1}. Please reflect on which of the variables/symbols in the equation are constants/parameters and which are actually dynamic variables. Then how they interact. This is the best I can do for you. I will be happy to get hired to do this work for you :-).
I hope this helps,
Marek
Re: discretize equation node for DBN
HI Marek;
If I understand what you mean, you said you could help me. Yes, I will be very happy if you can help me in this particular case as well. These are the equation data and this is the Bayesian network in question. I would be very grateful if you could check to see if the conditional probability table for node d can be obtained with the previous time step condition or not! just a tip about node K, i want to use ln(k) instead of k in equation with normal distribution with mean=2 and s.d=0.275
Thank you
If I understand what you mean, you said you could help me. Yes, I will be very happy if you can help me in this particular case as well. These are the equation data and this is the Bayesian network in question. I would be very grateful if you could check to see if the conditional probability table for node d can be obtained with the previous time step condition or not! just a tip about node K, i want to use ln(k) instead of k in equation with normal distribution with mean=2 and s.d=0.275
Thank you
-
- Site Admin
- Posts: 432
- Joined: Tue Dec 11, 2007 4:24 pm
Re: discretize equation node for DBN
Hi Bahman,
Sorry for the delay. As I suggested in my previous answer, we support our software but not really modeling effort, which requires domain knowledge and is always time consuming. We can do this for you on a consulting basis if this is something that would interest you.
Let me just answer your question about node d. The graph structure in your image shows that d_i is dependent on d_{i-1}, although the actual equation for d_i does not seem to be present in your image.
There is no problem with using logarithm in an equation in GeNIe -- you can use any function from among those listed on the right-hand side of the definition tab. At the moment, to use an equation node in a DBN, you have to discretize it first. We will integrate hybrid models with DBNs at some point but probably not in the next few months -- we have other important tasks ahead of us.
I hope this helps,
Marek
Sorry for the delay. As I suggested in my previous answer, we support our software but not really modeling effort, which requires domain knowledge and is always time consuming. We can do this for you on a consulting basis if this is something that would interest you.
Let me just answer your question about node d. The graph structure in your image shows that d_i is dependent on d_{i-1}, although the actual equation for d_i does not seem to be present in your image.
There is no problem with using logarithm in an equation in GeNIe -- you can use any function from among those listed on the right-hand side of the definition tab. At the moment, to use an equation node in a DBN, you have to discretize it first. We will integrate hybrid models with DBNs at some point but probably not in the next few months -- we have other important tasks ahead of us.
I hope this helps,
Marek