MAX doesn't work

The front end.
Post Reply
worldbreak
Posts: 3
Joined: Tue Nov 15, 2016 10:03 am

MAX doesn't work

Post by worldbreak »

I have a problem with using logical operators in Node Properties -> Definition. I have two nodes, which generates numbers from normal distribution. Then i have a node, which has to choose, which of these generated numbers are larger. I am using logical operator Max. But it doesn't work. When i directly say, that nodes are concrete numbers, it works well.

Can you help me, please?

Thank you so much.
marek [BayesFusion]
Site Admin
Posts: 430
Joined: Tue Dec 11, 2007 4:24 pm

Re: MAX doesn't work

Post by marek [BayesFusion] »

I'm not sure why you are saying that MAX does not work. I have just created a network as you described and tested it. I have run the standard number of samples and the child node (with the MAX function) seems to generate a correct distribution. I have changed the number of samples to 1 and the results seems to be correct as well (i.e., the child node picks correctly the larger of the two numbers. I may be misunderstanding your problem. Perhaps you can attach the network that you created?

Marek
worldbreak
Posts: 3
Joined: Tue Nov 15, 2016 10:03 am

Re: MAX doesn't work

Post by worldbreak »

I've attached the network now. The node doesn't working is C23. But maybe I have described the problem badly. Because there are two other nodes. I want to get the MAX value from C2 and C3.

Thank you so much
Attachments
example.xdsl
(2.76 KiB) Downloaded 441 times
marek [BayesFusion]
Site Admin
Posts: 430
Joined: Tue Dec 11, 2007 4:24 pm

Re: MAX doesn't work

Post by marek [BayesFusion] »

I have examined your model and I believe that the function is working fine. What it does is the following: For every sample generated for the nodes C2 and C3, it picks the higher of the two values as the sample for C23. What you see with the default number of samples (1000) is the distribution of samples at each of the nodes. To see that MAX is doing its job, please change the number of samples to 1 (Network Properties/Sampling) -- you will see this working for one sample. If you want to see it for each of the 1000 samples, you can always look at all the samples in the Value Tab of the nodes' properties. When you put the samples next to each other, you will see that the MAX function does precisely what I have described above :-).
I hope this helps.

Marek
worldbreak
Posts: 3
Joined: Tue Nov 15, 2016 10:03 am

Re: MAX doesn't work

Post by worldbreak »

I didn't know this.

Marek, thank you so much :-))).
FatmaNur
Posts: 5
Joined: Tue Dec 12, 2017 9:27 pm

Re: MAX doesn't work

Post by FatmaNur »

This discussion helped me a lot too! Is there any way we can compare the control values using MAX?
marek [BayesFusion]
Site Admin
Posts: 430
Joined: Tue Dec 11, 2007 4:24 pm

Re: MAX doesn't work

Post by marek [BayesFusion] »

Would you please clarify what you mean by "control values"?
FatmaNur
Posts: 5
Joined: Tue Dec 12, 2017 9:27 pm

Re: MAX doesn't work

Post by FatmaNur »

marek wrote:Would you please clarify what you mean by "control values"?
The node values that appear when the network is updated.
marek [BayesFusion]
Site Admin
Posts: 430
Joined: Tue Dec 11, 2007 4:24 pm

Re: MAX doesn't work

Post by marek [BayesFusion] »

I see. They are typically called "posterior probabilities." They are posterior probability distributions over the values of the of the node. Definition of a node refers to values of the parents and not probability distribution over these values, so the answer is I'm afraid negative -- there is no way to refer to elements of the distributions.
I hope this helps.

Marek
FatmaNur
Posts: 5
Joined: Tue Dec 12, 2017 9:27 pm

Re: MAX doesn't work

Post by FatmaNur »

marek wrote:I see. They are typically called "posterior probabilities." They are posterior probability distributions over the values of the of the node. Definition of a node refers to values of the parents and not probability distribution over these values, so the answer is I'm afraid negative -- there is no way to refer to elements of the distributions.
I hope this helps.

Marek
Thank you very much for your answer Marek. I will look into other ways to compare the posterior probabilities, such as exporting to Excel.
Post Reply