torch_concepts.data.PendulumDataset

class PendulumDataset(root: str | None = None, n_theta: int = 100, n_phi: int = 1000, seed: int = 42, concept_subset: list | None = None, label_descriptions: dict | None = None)

Procedurally generated pendulum scene dataset for regression.

Each sample is a rendered image of a pendulum with a light source casting a shadow. The concepts are the pendulum angle (theta) and the light angle (phi), both continuous. The regression task is to predict the x-coordinate of the pendulum ball.

Parameters:

• root (str, optional) – Root directory to store/load the dataset. If None, defaults to './data/pendulum'.

• n_theta (int, optional) – Number of theta angle steps for generation. Default: 100

• n_phi (int, optional) – Number of phi angle steps for generation. Default: 1000

• seed (int, optional) – Random seed for reproducibility. Default: 42

• concept_subset (list of str, optional) – Subset of concept names to use. Default: None (all concepts).

input_data

List of image filenames (images loaded on-the-fly).

Type:

list

concepts

Tensor of shape (n_samples, 3) containing [theta, phi, pendulum_x].

Type:

torch.Tensor

Examples

>>> from torch_concepts.data import PendulumDataset
>>> dataset = PendulumDataset(root='./data/pendulum', n_theta=10, n_phi=10)
>>> sample = dataset[0]
>>> x = sample['inputs']['x']  # image tensor (C, H, W)
>>> c = sample['concepts']['c']  # [theta, phi, pendulum_x]

__init__(root: str | None = None, n_theta: int = 100, n_phi: int = 1000, seed: int = 42, concept_subset: list | None = None, label_descriptions: dict | None = None)

Methods

__init__([root, n_theta, n_phi, seed, ...])
add_exogenous(name, value[, convert_precision])
add_scaler(key, scaler)	Add a scaler for preprocessing a specific tensor.
build()	Generate pendulum images and save metadata to disk.
collate(samples)	Collate samples into a batch, re-annotating the ground-truth concepts.
download()	This dataset is procedurally generated.
load()	Loads raw dataset and preprocess data.
load_raw()	Loads raw dataset without any data preprocessing.
maybe_build()
maybe_download()
remove_exogenous(name)
set_concepts(concepts)	Set concept annotations for the dataset.
set_graph(graph)	Set the adjacency matrix of the causal graph between concepts as a pandas DataFrame.

Attributes

annotations	Annotations for the concepts in the dataset.
concept_names	List of concept names in the dataset.
exogenous	Mapping of dataset's exogenous variables.
graph	Adjacency matrix of the causal graph between concepts.
has_concepts	Whether the dataset has concept annotations.
has_exogenous	Whether the dataset has exogenous information.
n_concepts	Number of concepts in the dataset.
n_exogenous	Number of exogenous variables in the dataset.
n_features	Shape of features in dataset's input (excluding number of samples).
n_samples	Number of samples in the dataset.
processed_filenames	The list of processed filenames in the self.root_dir folder that must be present in order to skip build().
processed_paths	The absolute paths of the processed files that must be present in order to skip building.
raw_filenames	The list of raw filenames in the self.root_dir folder that must be present in order to skip download().
raw_paths	The absolute paths of the raw files that must be present in order to skip downloading.
root_dir
shape	Shape of the input tensor.