/HybridLDA

Single machine version of LDA*

Primary LanguageJava

Single Machine version of LDA*

Compile

cd jni
make 
make jni

cd ..
mvn package

Run Hybrid LDA example

input=data/nips.train
K=1000
V=12420
threadNum=16
iterations=100
split=600
jvmOpts="-Djava.library.path=jni"
java -cp target/hybridlda-1.0-jar-with-dependencies.jar -Xmx1g $jvmOpts lda.parallel.CombineLDASeparate \
	$input $V $K $threadNum $iterations $split

where K is the number of topics, V is the number of word, split is the split point for document length. Documents with length less than split will be sampled by F+LDA while others will be sampled by WarpLDA.