recon_1prd_topn_par: Reconstruct predicate top scores on very large graphs

Description Usage Arguments Value Author(s) References See Also Examples

View source: R/recon_1prd_topn_par.R

Description

Reconstruct predicate top scores on very large graphs from RESCAL Factorization A & R. NB: A and R should be already loaded in the environment calling the function. Calculates top scores of A%*%R[[p]]%*%A^T. Uses chunks for rows and columns and constraints on maximum possible value of scores to avoid calculations of too small values

Usage

1
2
recon_1prd_topn_par(A,R,p, pcnt, rchLen = 1000, cchLen = 200, pve = 1e-10, 
mxrIter = 5, mxcIter = 25, grpLen = 40, OS_WIN = FALSE, dsname = "", ncores = 8)

Arguments

A

Embedding matrix part resulting from RESCAL factorization.

R

core tensor resulting from RESCAL factorization (r by r by m ).

p

predicate number.

pcnt

Number of triples in the predicate (to be reconstructed)

rchLen

row chunk length.

cchLen

column chunk length

pve

positive value: representing the smallest value of allowed score of reconstructed triple.

mxrIter

maximum number of iterations (chunks) in rows

mxcIter

maximum number of iterations (chunks) in columns

grpLen

length of one group of iterations, when running iterations in parallel results are collected for all iterations to be summarized after last iteration. Thus more memory is required. To avoid that iterations are divided to groups with summaries calculated for each group. Default 40.

ncores

number of cores used to run in parallel, 0 means no paralellism

OS_WIN

True when the operating system is windows, used to allow using Fork when running in parallel

dsname

optional:name of dataset

Value

The result is a LIST of three items:

ikv

A data frame containing the reconstructed triples (subject, Object, value, rchnk,chnk) using indexes of entities, rchnk and colun chunk are the chunk in which the triple is generated note predicate is an argument

minThr

the minimum score in each predicate (minimum score of a triple, threshold)

Iter

Number of iterations done

Author(s)

Abdelmoneim Amer Desouki

References

-Maximilian Nickel, Volker Tresp, Hans-Peter-Kriegel, "Factorizing YAGO: Scalable Machine Learning for Linked Data" WWW 2012, Lyon, France

-SynthG: mimicking RDF Graphs Using Tensor Factorization, Desouki et al. IEEE ICSC 2021

See Also

rescal scRescal RescalReconstructBack inv_rescal_sf_prd_chnkgrp

Examples

1
2
3
4
5
6
7
8
9
## Not run: 
#Reconstructing one predicate from DBpedia factorization
#print(load('dbpi42_r100.RData'))
lst=recon_1prd_topn_par(A=A,R=R,p,pcnt=prd_cnt1[p],rchLen=250000,cchLen=750,
                     mxrIter=30,mxcIter=70,ncores=12,grpLen=40,pve=pve,dsname=name)
    Res=cbind(S=lst$ikv[,1],P=p,lst$ikv[,2:5])#triples , value
 

## End(Not run)

RDFTensor documentation built on Jan. 16, 2021, 5:19 p.m.