search

MLDS-NUS / DeepLearningCustomThermodynamics

Code for reviewers
MIT License
7 stars 0 forks source link

Question about the implementation of Free-energy landscape analysis #1

Open jiali1025 opened 3 weeks ago

jiali1025 commented 3 weeks ago

Hi,

This is an excellent study, and I really love it.

However, I encountered some difficulties while trying to understand the code implementation for the Free-energy landscape analysis in your "potential" folder. It seems that the code primarily reproduces the plotting of V. While it’s easy to identify stable points where V = 0 from the diagram, I couldn’t find any code for identifying saddle points. I’m particularly interested in how you identified these points using BFGS. Additionally, I’d like to understand how you calculated the Jacobian matrix, along with its eigenvalues and eigenfunctions.

Many thanks in advance.

Kind regards, Jiali Li

jiali1025 commented 3 weeks ago

I am also quite curious about How to exploit automatic differentiation of neural networks and expand V(Z) in a Taylor series

chenxiaolichen commented 3 weeks ago

Hi,

This is an excellent study, and I really love it.

However, I encountered some difficulties while trying to understand the code implementation for the Free-energy landscape analysis in your "potential" folder. It seems that the code primarily reproduces the plotting of V. While it’s easy to identify stable points where V = 0 from the diagram, I couldn’t find any code for identifying saddle points. I’m particularly interested in how you identified these points using BFGS. Additionally, I’d like to understand how you calculated the Jacobian matrix, along with its eigenvalues and eigenfunctions.

Many thanks in advance.

Kind regards, Jiali Li

Hi Jiali, I put the code in the potential folder. You can check it there. Best, Xiaoli