parallelMF

Parallel implementation of Matrix Factorization using OpenMP. Given matrix R [mn] and k>0, solves R = XY where X [mk] and Y[k*n]

---

Requirements

This code requires openmp based c++ compiler. We have tested the code on crunchy machines with gcc-6.4.0.

---

Build

module load gcc-6.4.0

make

sample run:

./pmf -dim <dimension> -lambda <lambdaX> -mu <lambdaY> -iter <num-iter> -threads <threads for openmp> -lr <learning rate> -filename <datafile> -matSize <matrix size->m> <matrix size->n>

./pmf -dim 10 -lambda 0.01 -mu 0.01 -iter 10000 -threads 49 -lr 0.001 -filename data/1000/1000_1000_10-1.R -matSize 1000 1000

---

Directory structure

The source code is contained in the source/ dir. The matrix factorization code lives in pmf.h and the scheduler has been defined in the sched.h.

---