Hi,
I am a beginner of GeNIe. Now I have a sample datafile consisting of continuous & discrete datas. What I want to do is using 3 different discretized methods combining with 3 different algorithms (except naive) to get 9 different networks and test the accuracy of these networks respectively.
But when I conduct the discretized process of Hierarchical / Uniform-Width with learning method such as Essential graph search / PC with the same data file, there is no response and then the GeNie program is automatically closed. Could u pls explain why?
Sorry for my poor English. Thanks.
Apollo
doubt for learning new network
-
- Site Admin
- Posts: 1417
- Joined: Mon Nov 26, 2007 5:51 pm
Re: doubt for learning new network
GeNIe closing without any error message suggests a bug in the program. Can you post your discretized file (the one which causes GeNIe to disappear) here?hylx wrote:But when I conduct the discretized process of Hierarchical / Uniform-Width with learning method such as Essential graph search / PC with the same data file, there is no response and then the GeNie program is automatically closed. Could u pls explain why?
I am sorry, the data file is confidential so that I am afraid I cannot post here. I'm wondering if the problem is because I use wrong discretizing method: the sample data has 24 columns in which only one column has continuous data (high-moderate-low) and others are discrete. I know PC support both continuous and discrete data but they cannot both appear in the same file. But before I use PC to learn network, what I should do about discretizing the data? Discretize the datafile column by column which is discrete? I think I can transform the continuous column to discrete, but how to operate?
I don't know if u understand my point? Thanks so much.
I don't know if u understand my point? Thanks so much.
-
- Site Admin
- Posts: 1417
- Joined: Mon Nov 26, 2007 5:51 pm
Honestly, I'm not sure what you were trying to say in your last posthylx wrote:I don't know if u understand my point? Thanks so much.
If you have 24 columns, of which only one is continous, you can just discretize this single column and proceed with PC (because all data will be discrete).
With all discrete columns you can also try the GTT learning algorithm.
Finally, please note that we'll be releasing a build of GeNIe with completely new PC implementation shortly.
-
- Site Admin
- Posts: 1417
- Joined: Mon Nov 26, 2007 5:51 pm
If you mean that 'Discretize' command was disabled (grayed out) in the context menu, then it seems that GeNIe considers this column to be discrete. Can you copy/paste a couple of values from the column you want to discretize?hylx wrote:When I tried to discretize this single column from continuous to discrete one, I highlightened this column and right click but the discretize button is not activated.
I am faint now. The attachment is the part of datafile, isn't the first column a continuous and the second a discrete? BTW, the first column is the only "continuous" one as I said in the former post.
- Attachments
-
- dataset.jpg (7.45 KiB) Viewed 9320 times
Last edited by hylx on Fri May 23, 2008 6:16 pm, edited 1 time in total.
-
- Site Admin
- Posts: 1417
- Joined: Mon Nov 26, 2007 5:51 pm