The input file

Jump to navigation Jump to search

Brief descriptions

  • This file specifies some hyperparameters of the NEP potential [Fan2021].

File format

  • This file has the following fixed form:
cutoff              [cutoff_radial] [cutoff_angular]
n_max               [n_max_radial] [n_max_angular]
l_max               [l_max] 
number_of_neurons   [number_of_neurons]
regularization      [lambda_1] [lambda_2]
batch_size          [batch_size]
population_size     [population_size]
maximum_generation  [maximum_generation]
  • Do not modify the first column above.
  • For each row, items within [ ] are to be filled by the user.
  • We explain the meanings of the input items using the following example.
  • To fully understand this input file, one needs to consult [Fan2021].

An example

cutoff               8.0 4.0
n_max                12 8
l_max                6
number_of_neurons    25
regularization       0.01 0.05
batch_size           1000
population_size      50
maximum_generation   100000
  • Explanations:
    • The cutoff distance for the radial and angular descriptor components are [math]r_{\rm c}^{\rm R}=8.0[/math] A and [math]r_{\rm c}^{\rm A}=4.0[/math] A, respectively. We require that 1 A [math]\leq r_{\rm c}^{\rm A} \leq r_{\rm c}^{\rm R} \leq [/math] 10 A.
    • The Chebyshev polynomial expansion order for the radial and angular descriptor components are [math]n_{\rm max}^{\rm R}=12[/math] and [math]n_{\rm max}^{\rm A}=8[/math], respectively. We require that [math]0 \leq n_{\rm max}^{\rm R},n_{\rm max}^{\rm A} \leq 12 [/math].
    • The Legendre polynomial expansion order for the angular part is [math]l_{\rm max}=6[/math]. We require that [math]0 \leq l_{\rm max} \leq 6 [/math].
    • The number of neurons in the hidden layer (yes, we have tested that a single hidden layer is sufficient in NEP) is [math]N_{\rm neu}=25[/math]. We require that [math]1 \leq N_{\rm neu} \leq 50[/math].
    • The weight parameters for the [math]L_1[/math] and [math]L_2[/math] regularization are 0.01 and 0.05, respectively. These two parameters can take any non-negative values.
    • The batch size is [math]N_{\rm bat}=1000[/math], i.e., at each generation, 1000 structures will be evaluated. This parameter needs to be a positive integer. When it is equal to or larger than the total number of structures in the training set, it means using the full batch.
    • The population size is [math]N_{\rm pop}=50[/math]. We require that [math]10 \leq N_{\rm pop} \leq 100 [/math].
    • The training will be carried out for [math] N_{\rm gen}=10^5[/math] generations. We require that [math]10^2 \leq N_{\rm gen} \leq 10^6 [/math].


  • The cutoff distances [math]r_{\rm c}^{\rm R}[/math] and [math]r_{\rm c}^{\rm A}[/math] are in units of Angstrom.
  • Other quantities in this input file are dimensionless.


  • [Fan2021] Zheyong Fan, Zezhu Zeng, Cunzhi Zhang, Yanzhou Wang, Haikuan Dong, Yue Chen, and Tapio Ala-Nissila, Neuroevolution machine learning potentials: Combining high accuracy and low cost in atomistic simulations, To be submitted.