Skip to content

karmaresearch/statlearning

Repository files navigation

----------
Launching:
----------
Launch the prototype to create models with a <dataset> with 

1) TransE-SG.
	$ ./launcher_.sh launcher.py --fin <input/dataset.bin> --test-all 50 --nb 100 --me 300 --margin 2.0 --lr 0.1 --ncomp 50 --extvar True  --sampler=subgraph --resultsdir <path/to/resultDIR> --fout <path/to/save/models/file> --resultsfilename <outputfilename>

2) HOLE-SG:
	$ ./launcher_.sh launch_hole.py --fin <input/dataset.bin> --test-all 50 --nb 100 --me 300 --margin 0.2 --rparam 0.0 --lr 0.1 --ncomp 150 --extvar True  --sampler=subgraph --resultsdir <path/to/resultDIR> --fout <path/to/save/models/file> --resultsfilename <outputfilename>

--------------
Testing Model:
--------------

This runs with the parameters to run SRL models on given dataset.bin enabling creation of new variables (Subgraph Embeddings) with --extvar <True> option and also enbales negative sampling with --sampler=subgraph. To run Transe/Hole baseline make --extvar <False>, sampler=random-mode.

We also created a program to test each models, with various settings; this saves time from creating the models everytime. To test a model with;

	$./launcher_.sh <testmodel.py or testmodel_hole.py > --input <input/path/dataset.bin> --testValid TEST --testmode True --vectors <model> --fileTestValid <dataset/test.bin> --sampletest 1.0 --resultsdir <path/to/resultsDIR> --resultsfilename <outputfilename> --exvars True --post_threshold 5 --post2_enable True
	
To test a model with enriched dataset enable --exvars <True>. --sampletest <1.0> specifies the test data size (>=1.0) takes the full test data. The post processing parameters can be set here. 

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published