This is a package for tensor network renormalization group (TNRG) methods. It implements several 2D and 3D TNRG schemes, as well as useful procedures in TNRG calculations.
For 2D,
- Tensor Renormalization Group (TRG), not in its original form, but following Evenbly's implementation closely.
- Higher-order Tensor Renormalization Group (HOTRG). Instead of using the higher-order singular value decomposition, we use Evenlby's projective truncations described in this paper to formulate HOTRG and implement it.
- Evenly and Vidal's Tensor Network Renormalization (TNR), following Evenbly's implementation closely.
The first two schemes are different realizations of the block-tensor map in 2D. The last one is enhanced by entanglement filtering (or disentanglers) and goes beyond simple block-tensor map.
For 3D,
- Higher-order Tensor Renormalization Group (HOTRG). Again, we use Evenlby's projective truncations described in this paper to implement it.
- A HOTRG-like block-tensor map. It is similar to the HOTRG, but uses different isometric tensors for inner and outer legs, as is described in our preprint. It is more suitable than the HOTRG when an entanglement filtering process is incorporated.