Base classes

This section documents the abstract base classes that provide shared behavior across TINTOlib methods, following a scikit-learn style layout.

Inheritance overview

Method	Inherits from	Notes
TINTO	MappingMethod	Feature-to-pixel mapping + coordinates export
IGTD	MappingMethod	Feature-to-pixel mapping + coordinates export
REFINED	MappingMethod	Feature-to-pixel mapping + coordinates export
DeepInsight	ParamImageMethod	Parametric mapping (coords → pixels) + coordinates export
Fotomics	ParamImageMethod	Parametric mapping (coords → pixels) + coordinates export
BarGraph	AbstractImageMethod	Standard fit/transform utilities
DistanceMatrix	AbstractImageMethod	Standard fit/transform utilities
Combination	AbstractImageMethod	Standard fit/transform utilities
SuperTML	AbstractImageMethod	Standard fit/transform utilities
FeatureWrap	AbstractImageMethod	Standard fit/transform utilities
BIE	AbstractImageMethod	Standard fit/transform utilities

AbstractImageMethod

Abstract base class that defines the common training and image-generation pipeline for all methods.

Parameters

• problem : str, default=’classification’ The type of problem, defining how the images are grouped. Valid values: [‘classification’, ‘unsupervised’, ‘regression’].

• verbose : bool, default=False Whether to show execution details in the terminal.

• transformer : transformer or None, default=None Preprocessing transformation applied to features. Any scikit-learn compatible transformer or a custom implementation inheriting CustomTransformer.

• format : str, default=’png’ Output format using matplotlib images with [0,255] range for pixels or npy format.

Attributes

• problem : str Resolved problem type.

• verbose : bool Verbosity flag.

• transformer : transformer or None Preprocessing transformer.

• format : str Output format used by the method.

• _fitted : bool Indicates whether fit has been called.

Methods

• saveHyperparameters(filename=’objs.pkl’) Saves the configuration to disk as a pickle.

• loadHyperparameters(filename=’objs.pkl’) Loads a previously saved configuration.

• fit(data) Fits the model to tabular data.

• transform(data, folder) Generates and saves synthetic images.

• fit_transform(data, folder) Fits the model and generates images in one step.

Notes

• Validates input data (numeric only, no missing values).

• Automatically creates output folders and a CSV with image paths.

MappingMethod

Abstract base class for methods that map each feature to a pixel location. Inherits from AbstractImageMethod.

Parameters

• problem : str, default=’classification’

• verbose : bool, default=False

• transformer : transformer or None, default=None

• zoom : int, default=1 Multiplication factor for saving images.

• format : str, default=’png’

• cmap : str or None, default=None Matplotlib colormap. Use None for single-channel images.

Attributes

• zoom : int Zoom level for output images.

• cmap : str or None Colormap used by matplotlib.

• _features_mapping : pandas.DataFrame or None Feature-to-pixel mapping table.

Methods

• _get_features_mapping(features=None, columns_names=True) Returns feature positions in the image.

Notes

• Exports feature-to-pixel coordinates to features_positions.csv.

ParamImageMethod

Abstract base class for parametric mapping methods that first extract feature coordinates and then assign them to pixels. Inherits from MappingMethod.

Parameters

• dim : int Image dimension (square).

• problem : str, default=’classification’

• transformer : transformer or None, default=None

• verbose : bool, default=False

• assignment_method : str, default=’bin’ Technique to map features to pixels.

• relocate : bool, default=False Relocate features so each pixel represents a single feature.

• algorithm_opt : str, default=’linear_sum’ Optimization algorithm for assignment.

• group_method : str, default=’avg’ Strategy to compute pixel values when multiple features map to one pixel.

• zoom : int, default=1

• format : str, default=’png’

• cmap : str, default=’gray’

• random_seed : int, default=1

Attributes

• _image_dim : int Image dimension.

• _assignment_method : str

• _relocate : bool

• _algorithm_opt : str

• _group_method : str

• _random_seed : int

Notes

• Provides default pixel-value aggregation (average) and a relevance-weighted alternative when enabled.