Latent Dimensionality Reduction in Python
Project description
# PyLDR
Pyldr stands for (Python) Latent Dimensionality Reduction, and is a method for interpreting black box models. It is deployed here as a python module.
Black box models often provide better results than more interpretable methods, and brings some [quite strong opions](https://arxiv.org/abs/1811.10154). This method aims to bridge that gap by providing a generic, reliable algorithmic method for interpreting any model. I define interpretability as:
Understanding how a model understands the data, and whether it is similar to how a human would think of it.
Interpreting how the value of a feature, or subset of features, affects a model’s prediction (feature interpretation).
The ability to use a model when not all values for the input features are present.
## Running the Code
### Prerequesites
[Python3](https://www.python.org/download/releases/3.0/).
### Execution
All examples are contained in notebooks, while the LDR module is [ldr.py](ldr.py). The required packages are listed in [requirements.txt](requirements.txt), and their respective distributions and licenses can be found on the [Python Package Index](https://pypi.org/). To run the code use:
python3 -m pip install –requirement requirements.txt.
jupyter notebook.
The generated distribution example can be found [here](distribution_example.ipynb).
The classification example can be found [here](classification_example.ipynb).
The regression example can be found [here](regression_example.ipynb).
The step by step interpolation example can be found [here](interpolation_example.ipynb).
## Additional Notes
The [style sheet used](style.mplstyle) is from [one of my personal repos](https://github.com/Ekrekr/ekrekr.style).
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.