top of page
TensorTrace is an application designed to facilitate the implementation of tensor network algorithms 
TensorTrace provides a drag-and-drop interface for building tensor networks:
ss1temp.png
Completed networks can be exported from TensorTrace into code via copy-paste into a text editor:  
TT_MERA.png
TT_code.png
  • Flexible: code can be exported in the users choice of Python, Julia and MATLAB languages (where the Python code is also directly compatible with Google's TensorNetwork library).
  • Efficient: TensorTrace  contains built in solvers that automatically determine the optimal contraction order of each network.
Networks diagrams can be reconstructed by copy-pasting code back into TensorTrace:
TT_code2.png
TT_MPS.png
  • reconstruction of diagrams allows users to easily visualize and modify the networks associated to a computer code.
TensorTrace also provides a host of advanced features such as cost-scaling analysis and network auto-differentiaition:
der.png
der2.png
input network
single tensor derivatives
...
...
auto-differentiate
bottom of page