Scalable Tucker Factorization for Sparse Tensors - Algorithms and Discoveries (ICDE 2018)
Sejoon Oh, Namyong Park, Lee Sael, and U Kang
[Paper] [Supplementary Material]
Please refer to the following website for the details of P-Tucker (https://datalab.snu.ac.kr/ptucker/)
P-Tucker requires OpenMP 2.0 or above version! (if you use gcc/g++ compiler, it is installed by default, for macOS, see https://iscinumpy.gitlab.io/post/omp-on-high-sierra/ and replace Makefile with Makefile_mac)
"make" command will create a single executable file, which is "P-Tucker".
The executable file takes five arguments, which are the path of input tensor file, path of directory for storing results, tensor order, tensor rank, and number of threads. The arguments MUST BE valid and in the above order.
ex) ./P-Tucker input.txt result/ 3 10 20
If you put the command properly, P-Tucker will write all values of factor matrices and a core tensor in the result directory set by an argument. (PLEASE MAKE SURE THAT YOU HAVE A WRITE PERMISSION TO THE RESULT DIRECTORY!).
ex) result/FACTOR1, result/CORETENSOR
We note that input and output tensors are based on base-0 indexing, and tab- or space-separated input tensors are allowed.
To run the demo, please follow the following procedure. Sample tensor is created as 100x100x100 size with 1,000 observable entries.
1. Type "make demo"
2. Check "sample/result" directory for the demo factorization results
Orthogonalization of factor matrices and updating a core tensor are included in the source code (disabled by default). Please uncomment them (lines 471-472) before you run the code (if you wish to perform orthogonalization).
If you want to run P-Tucker-Cache or P-Tucker-APPROX, please contact the main author (Sejoon Oh, [email protected]).
We tested our proposed method P-Tucker in a Linux Ubuntu 16.04.3 LTS machine equipped with 20 Intel Xeon E5-2630 v4 2.2GHz CPUs and 512GB RAM.