emstoudenmire
commented
1 year ago Dear Qing, Thanks for your interest and glad you found the paper inspiring!
Unfortunately I don't have the information anymore about the library or operating system versions, since I wrote this code initially 7 years ago and haven't touched it since about 4 years ago. It is posted online purely for academic / educational purposes only, meaning just if one wants to read portions of the code. If you do want to get it to compile, the only thing you can do (which is what I would have to do also) is just to see what error messages you get and then fix the corresponding errors. Compilers such as clang can give very helpful error messages.
Also it's important to note that this code gives very far from state of the art performance. It was just my first attempt making a code to do this, and later others made faster codes. In general, the subject of using tensor networks to do machine learning is still very much a research subject and I don't think fast enough algorithms have really been discovered yet (gradient descent is in general a very slow algorithm, which is why so many machine learning groups have to use GPUs to get good enough results etc.). If you are interested in performing gradient descent on an MPS, though, there are more state of the art implementations or best of all you might want to write your own implementation.
Here is the message from the README: """ NOTE: these codes are research, proof-of-principle codes only, and are not intended to demonstrate the state of the art in terms of training times for matrix product states for machine learning
If you are seeking fast approaches for optimizing MPS, we recommend trying newer libraries which use stochastic gradient optimization methods, such as TorchMPS: https://github.com/jemisjoky/TorchMPS """
Good luck, and let me know if you have any technical questions about the paper.
Miles
QingXia0177
Dear Prof. Stoudenmire I am a Chinese postgraduate student majoring in quantum computing and machine learning. Recently, I read your paper ''Supervised Learning with Tensor Networks'', which inspired me a lot.But I encountered some issues that prevented the code from running successfully. I guess it's a version incompatibility issue. Now, I would like to know the version of itensor you are running this code with, and libpng version, png++version, gcc/g++version, and operating system version. I appreciate your help in advance. Best regards, Qing Xia