Problem Definition¶

This Jupyter Notebook demonstrates the usage of the ProblemDefinition class for defining machine learning problems using the PLAID library. It includes examples of:

Initializing a complete ProblemDefinition
Configuring problem characteristics and retrieve data
Saving and loading problem definitions

This notebook provides examples of using the ProblemDefinition class to define machine learning problems, configure characteristics, and save/load problem definitions.

Each section is documented and explained.

# Import required libraries
from pathlib import Path

import numpy as np

# Import necessary libraries and functions
from plaid import ProblemDefinition

Section 1: Initializing a ProblemDefinition¶

This section demonstrates how to initialize a ProblemDefinition and add input/output feature identifiers with the current API.

Initialize feature identifiers¶

scalar_1_feat_id = "Global/scalar_1"
scalar_2_feat_id = "Global/scalar_2"
scalar_3_feat_id = "Global/scalar_3"
field_1_feat_id = "Base_2_2/Zone/CellCenterFields/field_1"
field_2_feat_id = "Base_2_2/Zone/VertexFields/field_2"

print("#---# ProblemDefinition")
problem = ProblemDefinition(
    input_features=[scalar_3_feat_id, field_1_feat_id],
    output_features=[field_2_feat_id],
    train_split={"train": [0, 1]},
    test_split={"test": [2, 3]},
)
print(f"{problem = }")

#---# ProblemDefinition
problem = ProblemDefinition(input_features=['Base_2_2/Zone/CellCenterFields/field_1', 'Global/scalar_3'], output_features=['Base_2_2/Zone/VertexFields/field_2'], train_split={'train': [0, 1]}, test_split={'test': [2, 3]})

Add inputs / outputs to a Problem Definition¶

# Add unique input and output feature identifiers after initialization
#problem.add_input_features(scalar_1_feat_id)
#problem.add_output_features(scalar_2_feat_id)

# or Add list of input and output feature identifiers
problem.add_input_features([scalar_1_feat_id])
problem.add_output_features([scalar_2_feat_id])

print(f"{problem.input_features = }")
print(
    f"{problem.output_features = }",
)

problem.input_features = ['Base_2_2/Zone/CellCenterFields/field_1', 'Global/scalar_1', 'Global/scalar_3']
problem.output_features = ['Base_2_2/Zone/VertexFields/field_2', 'Global/scalar_2']

Section 2: Configuring Problem Characteristics and retrieve data¶

This section demonstrates how to handle and configure ProblemDefinition objects and access data.

Problem Definition identifier¶

ProblemDefinition has no stored name. When saved in a dataset, its identifier is the YAML filename stem or the key in a dict[str, ProblemDefinition].

Set Problem Definition split¶

# Current API uses required `train_split` and `test_split` fields.
# Note: each split field currently expects a dictionary with a single entry.
print(f"{problem.train_split = }")
print(f"{problem.test_split = }")

problem.train_split = {'train': [0, 1]}
problem.test_split = {'test': [2, 3]}

split_names = [next(iter(problem.train_split)), next(iter(problem.test_split))]
split_indices = {
    next(iter(problem.train_split)): next(iter(problem.train_split.values())),
    next(iter(problem.test_split)): next(iter(problem.test_split.values())),
}
print(f"{split_names = }")
print(f"{split_indices = }")

split_names = ['train', 'test']
split_indices = {'train': [0, 1], 'test': [2, 3]}

Show inputs / outputs¶

print(f"{problem.input_features = }")
print(f"{problem.output_features = }")

problem.input_features = ['Base_2_2/Zone/CellCenterFields/field_1', 'Global/scalar_1', 'Global/scalar_3']
problem.output_features = ['Base_2_2/Zone/VertexFields/field_2', 'Global/scalar_2']

Section 3: Saving and Loading Problem Definitions¶

This section demonstrates how to save and load a Problem Definition from a directory.

Save a Problem Definition to a YAML file¶

test_pth = Path(
    f"/tmp/test_safe_to_delete_{np.random.randint(low=1, high=2_000_000_000)}"
)
pb_def_save_fname = test_pth / "test_problem_definition.yaml"
test_pth.mkdir(parents=True, exist_ok=True)
print(f"saving path: {pb_def_save_fname}")

problem.save_to_file(pb_def_save_fname)

saving path: /tmp/test_safe_to_delete_1190088810/test_problem_definition.yaml

Load a ProblemDefinition from a YAML file¶

problem = ProblemDefinition.from_path(pb_def_save_fname)
print(problem)

input_features=['Base_2_2/Zone/CellCenterFields/field_1', 'Global/scalar_1', 'Global/scalar_3'] output_features=['Base_2_2/Zone/VertexFields/field_2', 'Global/scalar_2'] train_split={'train': [0, 1]} test_split={'test': [2, 3]}