DBN parameter learning from data set

[safaa]

hello;
thank you for your helpful answers.
to learn parameters there exist special format for variables to represent time (as it is shown in smile tutorials )
A A_0 A_1 B B_0 B_1 C C_0 C_1
t t t t t t t t t
t f f t f f t f f
what if we have 200 sample in each sample we have 20 events for example(from my data set which is about language processing

sample.txt

(5.74 KiB) Downloaded 347 times

the time series is word index in this case

[safaa]

I mean the data set is too large to re-order it. beside i couldn't understand how to deal with the time series dialog that shows up when I choose parameter learning.
I would be grateful for your answer.

[shooltz[BayesFusion]]

1. The column naming scheme which is used by GeNIe's automatic matching algorithm is A A_1 A_2 A_n ... B B_1 etc. Note that there's no "_t" suffix for t=0.

2. Any column names can be used for DBN parameter learning, but if they not follow the naming described above you'll have to manually setup the node/time/data column associations in the "Time series" dialog.

3. GeNIe won't be able to use data file structured like your sample.txt directly. You'll need to preprocess it to ensure each time step is represented by a separate column.

[safaa]

Thank you.