Skip to content

datahandler

Abstract Base Interface for Neural Network Data Handling in PDEs.

This module provides the base interface for handling data transformations and tensor conversions required for neural network-based PDE solvers. It defines the essential structure for managing various data types involved in finite element computations.

The implementation supports
  • Finite element data processing
  • Dirichlet boundary condition handling
  • Test point generation and management
  • Bilinear parameter tensor conversion
  • Sensor data generation and handling
  • Parameter management for inverse problems
Key classes
  • DataHandler: Abstract base class for PDE data handling
Note

All implementations assume double precision (float64) numpy arrays as inputs, with optional conversion to float32 for computational efficiency.

Dependencies
  • abc: For abstract base class functionality
  • tensorflow: For tensor operations
  • numpy: For numerical arrays
Authors
  • Thivin Anandh (https://thivinanandh.github.io/)
Versions
  • 27-Dec-2024 (Version 0.1): Initial Implementation

DataHandler

Abstract base class for PDE solution data handling and tensor conversion.

This class defines the interface for managing and converting various data types required in neural network-based PDE solvers. It handles transformation between numpy arrays and tensorflow tensors, and manages different aspects of the finite element computation data.

Attributes:

Name Type Description
fespace

Finite element space object containing mesh and element info
domain

Domain object containing geometric and boundary information
dtype

TensorFlow data type for tensor conversion (float32/float64)
Example

class MyDataHandler(DataHandler): ... def init(self, fespace, domain): ... super().init(fespace, domain, tf.float32) ... ... def get_dirichlet_input(self): ... # Implementation ... pass ... ... def get_test_points(self): ... # Implementation ... pass

Note

Concrete implementations must override: - get_dirichlet_input() - get_test_points() - get_bilinear_params_dict_as_tensors() - get_sensor_data() - get_inverse_params()

All methods should handle type conversion between numpy arrays and tensorflow tensors consistently.

Source code in scirex/core/sciml/fastvpinns/data/datahandler.py 
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
class DataHandler:
    """Abstract base class for PDE solution data handling and tensor conversion.

    This class defines the interface for managing and converting various data types
    required in neural network-based PDE solvers. It handles transformation between
    numpy arrays and tensorflow tensors, and manages different aspects of the finite
    element computation data.

    Attributes:
        fespace: Finite element space object containing mesh and element info
        domain: Domain object containing geometric and boundary information
        dtype: TensorFlow data type for tensor conversion (float32/float64)

    Example:
        >>> class MyDataHandler(DataHandler):
        ...     def __init__(self, fespace, domain):
        ...         super().__init__(fespace, domain, tf.float32)
        ...
        ...     def get_dirichlet_input(self):
        ...         # Implementation
        ...         pass
        ...
        ...     def get_test_points(self):
        ...         # Implementation
        ...         pass

    Note:
        Concrete implementations must override:
        - get_dirichlet_input()
        - get_test_points()
        - get_bilinear_params_dict_as_tensors()
        - get_sensor_data()
        - get_inverse_params()

        All methods should handle type conversion between numpy arrays
        and tensorflow tensors consistently.
    """

    def __init__(self, fespace, domain, dtype):
        """
        Constructor for the DataHandler class

        Args:
            fespace (FESpace2D): The FESpace2D object.
            domain (Domain2D): The Domain2D object.
            dtype (tf.DType): The tensorflow dtype to be used for all the tensors.

        Returns:
            None
        """

        self.fespace = fespace
        self.domain = domain
        self.dtype = dtype

    @abstractmethod
    def get_dirichlet_input(self) -> tuple:
        """
        This function will return the input for the Dirichlet boundary data

        Args:
            None

        Returns:
            The Dirichlet boundary data as a tuple of tensors
        """

    @abstractmethod
    def get_test_points(self):
        """
        Get the test points for the given domain.

        Args:
            None

        Returns:
            The test points as a tensor
        """

    @abstractmethod
    def get_bilinear_params_dict_as_tensors(self, function) -> dict:
        """
        Accepts a function from example file and converts all the values into tensors of the given dtype

        Args:
            function (function): The function from the example file which returns the bilinear parameters dictionary

        Returns:
            The bilinear parameters dictionary with all the values converted to tensors
        """

    @abstractmethod
    def get_sensor_data(self, exact_sol, num_sensor_points, mesh_type, file_name=None):
        """
        Accepts a function from example file and converts all the values into tensors of the given dtype

        Args:
            exact_sol (function): The exact solution function
            num_sensor_points (int): The number of sensor points
            mesh_type (str): The type of mesh
            file_name (str): The file name to save the sensor data

        Returns:
            The sensor data as a tensor
        """

    @abstractmethod
    def get_inverse_params(self, inverse_params_dict_function):
        """
        Accepts a function from example file and converts all the values into tensors of the given dtype

        Args:
            inverse_params_dict_function (function): The function from the example file which returns the inverse parameters dictionary

        Returns:
            The inverse parameters dictionary with all the values converted to tensors
        """

__init__(fespace, domain, dtype)

Constructor for the DataHandler class

Parameters:

Name Type Description Default
fespace FESpace2D

The FESpace2D object.

 required
domain Domain2D

The Domain2D object.

 required
dtype DType

The tensorflow dtype to be used for all the tensors.

 required

Returns:

Type Description

None
Source code in scirex/core/sciml/fastvpinns/data/datahandler.py 
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
def __init__(self, fespace, domain, dtype):
    """
    Constructor for the DataHandler class

    Args:
        fespace (FESpace2D): The FESpace2D object.
        domain (Domain2D): The Domain2D object.
        dtype (tf.DType): The tensorflow dtype to be used for all the tensors.

    Returns:
        None
    """

    self.fespace = fespace
    self.domain = domain
    self.dtype = dtype

get_bilinear_params_dict_as_tensors(function) abstractmethod

Accepts a function from example file and converts all the values into tensors of the given dtype

Parameters:

Name Type Description Default
function function

The function from the example file which returns the bilinear parameters dictionary

 required

Returns:

Type Description
dict

The bilinear parameters dictionary with all the values converted to tensors
Source code in scirex/core/sciml/fastvpinns/data/datahandler.py 
143
144
145
146
147
148
149
150
151
152
153
@abstractmethod
def get_bilinear_params_dict_as_tensors(self, function) -> dict:
    """
    Accepts a function from example file and converts all the values into tensors of the given dtype

    Args:
        function (function): The function from the example file which returns the bilinear parameters dictionary

    Returns:
        The bilinear parameters dictionary with all the values converted to tensors
    """

get_dirichlet_input() abstractmethod

This function will return the input for the Dirichlet boundary data

Returns:

Type Description
tuple

The Dirichlet boundary data as a tuple of tensors
Source code in scirex/core/sciml/fastvpinns/data/datahandler.py 
119
120
121
122
123
124
125
126
127
128
129
@abstractmethod
def get_dirichlet_input(self) -> tuple:
    """
    This function will return the input for the Dirichlet boundary data

    Args:
        None

    Returns:
        The Dirichlet boundary data as a tuple of tensors
    """

get_inverse_params(inverse_params_dict_function) abstractmethod

Accepts a function from example file and converts all the values into tensors of the given dtype

Parameters:

Name Type Description Default
inverse_params_dict_function function

The function from the example file which returns the inverse parameters dictionary

 required

Returns:

Type Description

The inverse parameters dictionary with all the values converted to tensors
Source code in scirex/core/sciml/fastvpinns/data/datahandler.py 
170
171
172
173
174
175
176
177
178
179
180
@abstractmethod
def get_inverse_params(self, inverse_params_dict_function):
    """
    Accepts a function from example file and converts all the values into tensors of the given dtype

    Args:
        inverse_params_dict_function (function): The function from the example file which returns the inverse parameters dictionary

    Returns:
        The inverse parameters dictionary with all the values converted to tensors
    """

get_sensor_data(exact_sol, num_sensor_points, mesh_type, file_name=None) abstractmethod

Accepts a function from example file and converts all the values into tensors of the given dtype

Parameters:

Name Type Description Default
exact_sol function

The exact solution function

 required
num_sensor_points int

The number of sensor points

 required
mesh_type str

The type of mesh

 required
file_name str

The file name to save the sensor data

 None

Returns:

Type Description

The sensor data as a tensor
Source code in scirex/core/sciml/fastvpinns/data/datahandler.py 
155
156
157
158
159
160
161
162
163
164
165
166
167
168
@abstractmethod
def get_sensor_data(self, exact_sol, num_sensor_points, mesh_type, file_name=None):
    """
    Accepts a function from example file and converts all the values into tensors of the given dtype

    Args:
        exact_sol (function): The exact solution function
        num_sensor_points (int): The number of sensor points
        mesh_type (str): The type of mesh
        file_name (str): The file name to save the sensor data

    Returns:
        The sensor data as a tensor
    """

get_test_points() abstractmethod

Get the test points for the given domain.

Returns:

Type Description

The test points as a tensor
Source code in scirex/core/sciml/fastvpinns/data/datahandler.py 
131
132
133
134
135
136
137
138
139
140
141
@abstractmethod
def get_test_points(self):
    """
    Get the test points for the given domain.

    Args:
        None

    Returns:
        The test points as a tensor
    """