API¶
This is the full API documentation of the research package.
research.active_learning¶
This submodule contains the code developed for experiments related to Active Learning.
|
Class to simulate Active Learning experiments. |
research.data_augmentation¶
Contains the implementation of variations of oversampling/data augmentation algorithms, as well as helper classes to use oversampling algorithms as data augmentation techniques.
Class to to perform over-sampling using Geometric SMOTE. |
|
A wrapper to facilitate the use of imblearn.over_sampling objects for data augmentation. |
research.datasets¶
Download, transform and simulate various datasets.
These classes were extracted from the utils.py script from AlgoWit’s publications repo, to which I have also contributed.
Link to related repo: https://github.com/AlgoWit/publications
|
Base class to download and save datasets. |
|
Class to download, transform and save binary class datasets. |
|
Class to download, transform and save binary class imbalanced datasets. |
Class to download, transform and save datasets with both continuous and categorical features. |
|
|
Class to download, transform and save multiclass datasets. |
|
Class to download, transform and save remote sensing datasets. |
research.metrics¶
This submodule contains various performance metrics/scorers that are not included in scikit-learn’s scorers’ dictionary. Additionally, an expanded dictionary of scorers (as compared with scikit-learn’s) is also provided.
Parts of this code was taken from the utils.py script from AlgoWit’s publications repo, to which I have also contributed.
Link to related repo: https://github.com/AlgoWit/publications
|
Geometric mean score with macro average. |
Area under the learning curve. |
|
|
Data Utilization Rate. |
|
Make an Active Learning scorer from a AL-specific metric or loss function. |
research.utils¶
This submodule contains a variety of general utility functions as well as tools used to format and prepare tables to incorporate into LaTeX code.
Additionally, an expanded (as compared with scikit-learn’s) dictionary of scorers is also provided.
This code was taken from the utils.py script from AlgoWit’s publications repo, to which I have also contributed.
Link to related repo: https://github.com/AlgoWit/publications
|
Generate table that combines mean and sem values. |
Format p-values. |
|
|
Sort tables rows and columns. |
|
Generate data, results and analysis paths. |
|
Make bold the lowest or highest value(s). |
|
Generate table that combines mean and sem values. |
Converts an image as numpy array (with ground truth) to a pandas dataframe |
|
|
Load datasets from sqlite database and/or csv files. |
|
Extract estimators and parameters grids. |
|
|
|
Load LaTeX style configurations for Matplotlib/Seaborn Visualizations. |
|
Converts a column of values to hex-type colors. |