8000 Error with approximate counting (FACT algorithm) · Issue #37 · starling-lab/BoostSRL · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

Error with approximate counting (FACT algorithm) #37

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
CheongWoong opened this issue May 29, 2020 · 2 comments
Open

Error with approximate counting (FACT algorithm) #37

CheongWoong opened this issue May 29, 2020 · 2 comments

Comments

@CheongWoong
Copy link
CheongWoong commented May 29, 2020 •
edited
Loading

I tried to experiment with approximate counting (FACT algorithm) using the commands below.

java -cp boostsrl_v1.1.1.jar edu.wisc.cs.will.Boosting.MLN.RunBoostedMLN -l -train train/ -target (target) -model model/ -trees 10 -mln -approxCount
java -cp boostsrl_v1.1.1.jar edu.wisc.cs.will.Boosting.MLN.RunBoostedMLN -i -test test/ -target (target) -model model/ -trees 10 -mln -approxCount -aucJarPath .

I experimented on four datasets: UW-CSE, WebKB, NELL-Sports, CiteSeer, using the exactly same data from this repository. (The same mode, fact and target)

For CiteSeer, I encountered the following error while learning.
image

For the other datasets, learning and inference are done without errors. However, the result seems wrong (every prediction has the same probability as shown below), which is different from the result reported in this paper (Fast Relational Probabilistic Inference and Learning: Approximate Counting via Hypergraphs).
image

I used boostsrl v1.1.1 with java 14.0.1 on Linux. MLN_Boost (without -approxCount flag) works fine without any errors.

(1) How can I resolve the error that I encountered on CiteSeer?
(2) How can I reproduce the experimental results reported in Fast Relational Probabilistic Inference and Learning: Approximate Counting via Hypergraphs?
image

Thank you in advance,
-Cheongwoong Kang
@nandhiniramanan5
Copy link
Contributor
nandhiniramanan5 commented May 29, 2020 •
edited
Loading

Hi Kang,

Thanks a lot for pointing this out. If you are willing to run the code for the paper "Fast Relational Probabilistic Inference and Learning: Approximate Counting via Hypergraphs", please look at the following repository for the code of MACH algorithm.
https://github.com/mayukhdas/MACH.git

We will push this code into the main code repo (starling.github) soon. Meanwhile, please find the code and instructions in the above link. Let us know if you need any more assistance.

Regards,
Nandini Ramanan

@CheongWoong
Copy link
Author

Hi Ramanan,

Thanks for taking time for this.

However, I am not running the code for MACH algorithm. I want to run the code for FACT algorithm and reproduce the result of FACT algorithm reported in "Fast Relational Probabilistic Inference and Learning: Approximate Counting via Hypergraphs".

However, I encountered errors and wrong prediction results when I use -approxCount flag as mentioned above.
Please consider taking a look at the errors that I mentioned above.

Thank you,
Cheongwoong Kang

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@nandhiniramanan5 @CheongWoong
0