Neural Network Inquiry #31
-
|
Hello, I hope you are doing well. I was reviewing the Colab tutorial and the example for Silicon has intrigued me. The AIMD dataset, upon which the neural network is trained, states to have been sampled for T=800K with 64 Si atoms. The resultant plot after the LAMMPS simulation is an rdf distribution for T=300K. How can this be if the input data does not cover 300K? Apologies if I am misunderstanding something. Thank you! |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
|
Machine Learning Interatomic Potentials generalize well from high-temperature to low-temperature data. Data sampled at high-T cover a broad configuration space of diverse structures, from which it is often possible to generalize to lower-T data. |
Beta Was this translation helpful? Give feedback.
-
|
Got it, thank you. |
Beta Was this translation helpful? Give feedback.
Machine Learning Interatomic Potentials generalize well from high-temperature to low-temperature data. Data sampled at high-T cover a broad configuration space of diverse structures, from which it is often possible to generalize to lower-T data.