BIGtensor

Overview

BIGtensor is a large-scale tensor mining package. Users can select types to run the BIGtensor. BIGtensor (GPU) is designed to fully utilize CPU and GPUs. BIGtensor (Hadoop) also runs on a distributed Hadoop platform, an open-source version of MapReduce framework.

BIGtensor provides various tensor operations:

• Tensor decompositions: PARAFAC, Nonnegative PARAFAC, Tucker, Nonnegative Tucker, Coupled Matrix-tensor Factorization (Tucker),
• Tensor generations: generation where all tensor values are set to 1, random tensor generation, tensor generation from factor matrices and core tensor
• Tensor-Tensor operation: n-mode product, scalar operations, etc.
• Tensor manipulation (operation): matricization, convertsign, collapse, permutation, etc.

Contributors

BIGtensor is develeloped in Data Mining Lab @ Seoul National University.
The main contributors are:

Press

BIGtensor receives wide media coverage.

• YONHAP NEWS (in Korean)