pinnx.utils module

Internal utilities.

This module provides utility functions for array operations, data conversion, visualization, sampling, and other helper functions used throughout PINNx.

Conversion Helpers

Functions for converting between dictionary and array representations.

array_to_dict	Convert args to a dictionary.
dict_to_array	Convert a dictionary to an array.

Array Operations

Utilities for tensor and array manipulation.

is_tensor
istensorlist
convert_to_array	Convert a list of numpy arrays or tensors to a numpy array or a tensor.
hstack
zero_padding

Training Utilities

Functions for training, visualization, and result management.

apply: Apply a function element-wise.

standardize: Standardize data to zero mean and unit variance.

saveplot: Save and plot training results.

plot_loss_history: Plot loss history over iterations.

save_loss_history: Save loss history to file.

plot_best_state: Plot best model state.

save_best_state: Save best model state.

dat_to_csv: Convert .dat files to .csv format.

isclose: Check if values are close within tolerance.

smart_numpy: Convert to NumPy array intelligently.

apply	Launch a new process to call the function.
standardize	Standardize features by removing the mean and scaling to unit variance.
saveplot	Save/plot the loss history and best trained result.
plot_loss_history	Plot the training and testing loss history.
save_loss_history	Save the training and testing loss history to a file.
plot_best_state	Plot the best result of the smallest training loss.
save_best_state	Save the best result of the smallest training loss to a file.
dat_to_csv	Converts a dat file to CSV format and saves it.
isclose	A modified version of np.isclose for DeepXDE.
smart_numpy

Internal Helpers

Internal utility functions for PINNx operations.

timing	Decorator for measuring the execution time of methods.
merge_dict
subdict	Get a sub-dictionary from a dictionary.
check_not_none
run_if_all_none
run_if_any_none
vectorize	numpy.vectorize wrapper that works with instance methods.
return_tensor	Convert the output to a Tensor.
to_numpy	Create numpy ndarrays that shares the same underlying storage, if possible.
make_dict	Convert two lists or two variables into a dictionary.
save_animation
list_to_str

Loss Functions

Loss functions for training neural networks.

mean_absolute_error
mean_squared_error
mean_l2_relative_error
softmax_cross_entropy
get_loss	Retrieves a loss function.

Sampling Helpers

Batch sampling utilities.

BatchSampler

Samples a mini-batch of indices.

Sampling Strategies

Point sampling strategies for training data generation.

sample: Generic sampling function.

pseudorandom: Pseudorandom sampling.

quasirandom: Quasi-random (low-discrepancy) sampling.

InitialPointGenerator: Generates initial training points.

check_random_state: Validates random state.

check_dimension: Validates dimension parameter.

sample	Generate pseudorandom or quasirandom samples in [0, 1]^dimension.
pseudorandom	Pseudo random.
quasirandom
InitialPointGenerator
check_random_state
check_dimension	Turn a provided dimension description into a dimension object.

Transformers

Data transformation and preprocessing utilities.

Transformer: Base transformer class.

Identity: Identity transformation.

StringEncoder: Encode strings to numerical values.

LogN: Logarithmic transformation.

CategoricalEncoder: Encode categorical variables.

LabelEncoder: Encode labels.

Normalize: Normalize data.

Pipeline: Chain multiple transformers.

Transformer	Base class for all 1-D transformers.
Identity	Identity transform.
StringEncoder	StringEncoder transform.
LogN	Base N logarithm transform.
CategoricalEncoder	OneHotEncoder that can handle categorical variables.
LabelEncoder	LabelEncoder that can handle categorical variables.
Normalize	Scales each dimension into the interval [0, 1].
Pipeline	A lightweight pipeline to chain transformers.