SymmetryFunctionDataset

class hdnnpy.dataset.descriptor.symmetry_function_dataset.SymmetryFunctionDataset(order, structures, **func_param_map)[source]

Bases: hdnnpy.dataset.descriptor.descriptor_dataset_base.DescriptorDatasetBase

Symmetry function dataset for descriptor of HDNNP.

It accepts 0 or 2 for order.

Each symmetry function requires following parameters.
Pass parameters you want to use for the dataset as keyword arguments func_param_map.
  • type1: \(R_c\)
  • type2: \(R_c, \eta, R_s\)
  • type4: \(R_c, \eta, \lambda, \zeta\)
Parameters:
  • order (int) – passed to super class.
  • structures (list [AtomicStructure]) – passed to super class.
  • **func_param_map (list [tuple]) – parameter sets for each type of symmetry function.

References

Symmetry function was proposed by Behler et al. in this paper as a descriptor of HDNNP. Please see here for details of each symmetry function.

__getitem__(item)

Return descriptor data this instance has.

If item is string, it returns corresponding descriptor. Available keys can be obtained by descriptors attribute. Otherwise, it returns a list of descriptor sliced by item.

__len__()

Number of atomic structures given at initialization.

calculate_descriptors(structure)[source]

Calculate required descriptors for a structure data.

Parameters:structure (AtomicStructure) – A structure data to calculate descriptors.
Returns:Calculated descriptors. The length is the same as order given at initialization.
Return type:list [ndarray]
clear()

Clear up instance variables to initial state.

differentiate()[source]

Decorator function to differentiate symmetry function.

generate_feature_keys(elements)[source]

Generate feature keys from given elements and parameters.

parameters given at initialization are used.
This method is used to initialize instance and expand feature dimension in HDNNPDataset.
Parameters:elements (list [str]) – Unique list of elements. It should be sorted alphabetically.
Returns:Generated feature keys in a format like <func_name>:<parameters>:<elements>.
Return type:list [str]
load(file_path, verbose=True, remake=False)

Load dataset from .npz format file.

Only root MPI process load dataset.

It validates following compatibility between loaded dataset and atomic structures given at initialization.

  • length of data
  • elemental composition
  • elements
  • tag

It also validates that loaded dataset satisfies requirements.

  • feature keys
  • order
Parameters:
  • file_path (Path) – File path to load dataset.
  • verbose (bool, optional) – Print log to stdout.
  • remake (bool, optional) – If loaded dataset is lacking in any feature key or any descriptor, recalculate dataset from scratch and overwrite it to file_path. Otherwise, it raises ValueError.
Raises:
  • AssertionError – If loaded dataset is incompatible with atomic structures given at initialization.
  • ValueError – If loaded dataset is lacking in any feature key or any descriptor and remake=False.
make(verbose=True)

Calculate & retain descriptor dataset

It calculates descriptor dataset by data-parallel using MPI communication.
The calculated dataset is retained in only root MPI process.
Parameters:verbose (bool, optional) – Print log to stdout.
save(file_path, verbose=True)

Save dataset to .npz format file.

Only root MPI process save dataset.

Parameters:
  • file_path (Path) – File path to save dataset.
  • verbose (bool, optional) – Print log to stdout.
Raises:

RuntimeError – If this instance do not have any data.
DESCRIPTORS = ['sym_func', 'derivative', 'second_derivative']

Names of descriptors for each derivative order.

Type:list [str]
descriptors

Names of descriptors this instance have.

Type:list [str]
elemental_composition

Elemental composition of atomic structures given at initialization.

Type:list [str]
elements

Elements of atomic structures given at initialization.

Type:list [str]
feature_keys

Unique keys of feature dimension.

Type:list [str]
function_names

Names of symmetry functions this instance calculates or has calculated.

Type:list [str]
has_data

True if success to load or make dataset, False otherwise.

Type:bool
n_feature

Length of feature dimension.

Type:int
name = 'symmetry_function'

Name of this descriptor class.

Type:str
order

Derivative order of descriptor to calculate.

Type:int
params

Mapping from symmetry function name to its parameters.

Type:dict [list [tuple]]]
tag

Unique tag of atomic structures given at initialization.

Usually, it is a form like <any prefix> <chemical formula>. (ex. CrystalGa2N2)

Type:str