Learn New Network... Method... continuous data only?

The front end.
Post Reply
sykes
Posts: 10
Joined: Tue Jan 15, 2008 8:03 pm

Learn New Network... Method... continuous data only?

Post by sykes »

Hi,

I am trying to use Data->Learn New Network
with a datafile that has only numeric data (some columns are integers, while others are floats).

In the tutorial #15 under Learning Structure it says:
"Note that if the data set contains continuous variables they will need to be discretized for some learning methods to be able to run, e.g. Naive."

But I keep getting:
"Selected algorithm doesn't support continuous columns."
for all 4 of the available Methods...PC, Naive, etc.

i hope there is a solution/work around because I would like to keep the data in its original continuous form (i.e, I don't want to discretize it).

Thanks for your help,

Sincerely,
Ed Sykes
shooltz[BayesFusion]
Site Admin
Posts: 1417
Joined: Mon Nov 26, 2007 5:51 pm

Re: Learn New Network... Method... continuous data only?

Post by shooltz[BayesFusion] »

sykes wrote:But I keep getting:
"Selected algorithm doesn't support continuous columns."
for all 4 of the available Methods...PC, Naive, etc.
PC supports continuous data now; the learned network will contain equation-based nodes. If you want to create discrete network from continuous data you'll need to discretize the data first or discretize the equation-based network after learning.
i hope there is a solution/work around because I would like to keep the data in its original continuous form (i.e, I don't want to discretize it).
If you save your data as .gdat file from GeNIe, the original continuous values will be preserved, even if discretization is applied. You'll be able to disable discretization or change the intervals later.
sykes
Posts: 10
Joined: Tue Jan 15, 2008 8:03 pm

Post by sykes »

Dear Shooltz,

thanks for your help. I've downloaded the most recent version and have saved the same dataset (containing only continuous data...reals and integers) as a .gdat file.

I am trying to use PC as the method...here is the error message:

"Selected algorithm supports continuous or discrete columns, but they can't be used together."

I've gone through my dataset and can't find any "discrete" columns...I tried attaching the file, but I keep getting "Tried to upload empty file"
.... could you advise on how I could locate where GeNIe is confused about my dataset?

Thanks so much,
Ed Sykes
shooltz[BayesFusion]
Site Admin
Posts: 1417
Joined: Mon Nov 26, 2007 5:51 pm

Post by shooltz[BayesFusion] »

sykes wrote:I've gone through my dataset and can't find any "discrete" columns...I tried attaching the file, but I keep getting "Tried to upload empty file"
Try attaching zipped file.

.... could you advise on how I could locate where GeNIe is confused about my dataset?
GeNIe considers the data column to be continuous if all values are numeric and a) at least one value is non-integer or b) distinct number of values is greater than 'Discrete threshold' parameter which you can specify in the 'Learn New Network' dialog (the default value is 20).

Maybe one or more of your columns is all-integer, but has less than 20 distinct numbers?
sykes
Posts: 10
Joined: Tue Jan 15, 2008 8:03 pm

Post by sykes »

Dear Shooltz,

Thanks for your help. You are right. The problem was my integer columns. It works now.

Cheers,
Ed
Post Reply