distxy

distxy is a Python library designed to calculate the distance between two probability distributions. It offers a variety of distance and divergence measures, allowing users to quantify how different two probability distributions are. The library supports both discrete and continuous distributions, though it primarily focuses on discrete cases where distributions are represented as collections of probabilities for specific outcomes.

The core functionality of distxy lies in its implementation of various metrics. These include, but are not limited to, the Kullback-Leibler divergence (KL divergence), Jensen-Shannon divergence (JS divergence), total variation distance, and Hellinger distance. Each of these measures provides a different perspective on the dissimilarity between distributions. For instance, KL divergence is asymmetric and measures how much information is lost when one distribution is used to approximate another, while JS divergence is symmetric and is a smoothed version of KL divergence.

Users typically provide two distributions to the library, often represented as lists, NumPy arrays, or dictionaries where keys are outcomes and values are their corresponding probabilities. The library then computes the chosen distance metric between these inputs. This makes distxy useful in various fields such as machine learning, statistical modeling, and information theory, where comparing probability distributions is a common task. For example, in model evaluation, one might use distxy to compare the distribution of predicted outcomes with the distribution of actual outcomes.