Hello,
I am trying to use the equation nodes to compute reliabilities of very reliable information systems. This means that we get numbers with a very high number of decimals (0,9999904632 etc.). We want to use these numbers in products.
It appears GeNIe only allows six figures in the computations which introduces many rounding errors into our computations. Is there any way to change the size of the numbers so that we could have bigger numbers with more decimals? Or, is it possible to increase the precision in the SMILE API?
Best regards
Per
Equation node precision
-
- Site Admin
- Posts: 1417
- Joined: Mon Nov 26, 2007 5:51 pm
Re: Equation node precision
Most of the equation-based inference algorithms are approximate, so even with increased decimal count you'd probably be unable to get reasonable results.Per Närman wrote:It appears GeNIe only allows six figures in the computations which introduces many rounding errors into our computations. Is there any way to change the size of the numbers so that we could have bigger numbers with more decimals? Or, is it possible to increase the precision in the SMILE API?
There's no method of increasing the precision in the API at this point.
-
- Posts: 4
- Joined: Mon May 09, 2011 1:44 pm
Re: Equation node precision
Hi,
Can you tell how can we connect an arc node to other types of nodes.
Thanks
Can you tell how can we connect an arc node to other types of nodes.
Thanks
-
- Site Admin
- Posts: 1417
- Joined: Mon Nov 26, 2007 5:51 pm
Re: Equation node precision
Did you mean the equation (not 'arc') nodes? They can only be linked to other equation nodes.gingerauto wrote:Hi,
Can you tell how can we connect an arc node to other types of nodes.
-
- Posts: 4
- Joined: Mon May 09, 2011 1:44 pm
Re: Equation node precision
hi,
Yes i mean equation node..we can't, but how can we represent a node which follows for example an exponential or normal probability law...
Thanks
Yes i mean equation node..we can't, but how can we represent a node which follows for example an exponential or normal probability law...
Thanks