Reconstruction of TTNs

We can also optimize the network structure given a tree tensor network using the TTNOpt package.

We assume that the input is given as a TTN, local tensors and the .dat file that specify the connectivity of the network. All input files should be included in the same directory such as:

├── data
│   ├──edges.dat
│   ├──isometry0.npy
│   ├──isometry1.npy
│   ├──isometry2.npy
│   ├──isometry3.npy
...

A input file for the reconstruction should be written as follows:

reconstruction.yml

target:
    dir: data
    tensors_name: isometry
    graph_file: edges.dat

numerics:
    opt_structure:
        type: 1 # 0: no optimization, 1: structural optimization
    max_num_sweep: 20
    truncation_error: 1e-8

output:
    dir: outout

Then, we run the reconstruction by the following command:

ft reconstruction.yml

The output will be saved in the directory specified in the output section of the input file.

Examples of the input files are provided in the sample directory. Especially, by running the run_reconstruction.sh script, we can reproduce the results of the paper, that is, the reconstruction of the multivariate normal distribution.