Getting Started
Getting Started
Download
The software has recently moved to GITLAB: http://gitlab.com/tensors/tensor_toolbox. Official releases can be obtained by visiting the Releases page. The latest development realease can be obtained by cloning or downloading from the main GITLAB page. Version 2.6 and earlier can be obtained here.
Installation
- Unpack the files, if necessary
- Start MATLAB
- Within MATLAB, navigate to the tensor_toolbox
directory and execute the following commands:
- addpath(pwd)
- savepath
Getting help
At any time, type help tensor_toolbox for help on classes or functions. You can also find a getting started guide via MATLAB's help system. Launch help by pressing the question mark button and look for Tensor Toolbox under supplemental software, as highlighted in the image below.
What's new in Version 3.1?
Version 3.1 adds
- New location on GITLAB plus open source BSD licence.
- New classes and functions for symmetric tensors: symtensor, symktensor, cp_sym, tucker_sym
- New class for sums of different tensor types: sumtensor
- Function to compute HOSVD and ST-HOSVD: hosvd
- New function for CP with alternating randomized least squares(also known as CPRAND): cp_arls
- New generalized CP, including stochastic version: gcp_opt
- New method for creating binary problems where the nonnegative model tensor represents the odds of seeing a one: create_problem_binary
- Added methods to create all-ones sparse tensor that has the same pattern as a given sparse tensor: sptensor/spones
- New function to vizualize CP-decompoised Kruskal tensors: ktensor/viz
We have also fixed many bugs. View the RELEASE_NOTES.txt file for details.