1: \begin{abstract}

2: We revisit the corner transfer matrix renormalization group (CTMRG) method of Nishino and Okunishi 

3: for contracting two-dimensional (2D) tensor networks and demonstrate that its performance can be 

4: substantially improved by determining the tensors using an eigenvalue solver as opposed to the 

5: power method used in CTMRG. 

6: We also generalize the variational uniform matrix product state (VUMPS) ansatz for 

7: diagonalizing 1D quantum Hamiltonians to the case of 2D transfer matrices and discuss similarities 

8: with the corner methods. 

9: These two new algorithms will be crucial to improving the performance of variational infinite 

10: projected entangled pair state (PEPS) methods. 

11: \end{abstract}

12: