pepperon.ai is an open-source Python toolset for machine learning and data science and is distributed under the MIT License (Please don't sue me).

This project was started in 2019 by @JonWiggins as a central repository for some of the fun algorithms from various courses at the University of Utah.

It is intended to be a lot like scikit-learn, except more buggy, with less functionality, worse documentation, and not used by anyone. But that's okay, because making it will be fun.

pepperon.ai is built on a few core ideas:

1. Machine Learning should be accessable to the masses
2. Bugs should be more common in software packages
3. Runtime, much like digging holes, builds character

pepperon.ai is currently on version 0.12.0

pepperon.ai requires:

• Python (>= 3.5)
• NumPy (>= 1.11.0)
• A lot of Patience
• Scipy (>= 0.11)
• Pandas (>= 0.24)

You can install pepperonai from pypi via:

pip install pepperonai

• Decision Trees created with ID3 ; for all your decison needs

• Simple Perceptron ; for all your needs that are both linearly seperable and basic
• Average Perceptron ; for all your needs that are both linearly seperable and noisy

• SVM on SGD ; for when you are using Perceptron, and decide you want it to be better

• Guassian and Bernoulli

• Random Forest on ID3

• Labeling data based on it's nearest neighbors

• For line based modeling

• ngrams ; for all your simplistic corpus needs

• kgrams ; for all your shingling needs

• For creating a heavy hitters lower bound

• For creating a heavy hitters upper bound

• In progress
• Fowlkes-Mallows Index ; for comparing clusterings;
• Purity Index ; for comparing clusterings
• Grid Search for finding the best hyperparameters for your model
• Jackknifing to create cross validation folds
• General testing and evaluating methods

Clusters based on the kmeans++ algorithm, all based on probability

Clusters based on the Greedy Gonlzales algorithm, iteratively picks the furthest point from the existing clusters to be a new center.

Rather than basing off of clusters, Heirarchial methods iteratively merge the two nearest clusters. What defines near is based on the linking method given, the built in linking functions provided are:

• Single link: finds the smallest distance between two points in clusters
• Complete link: find the largest distance between two points in clusters
• Mean link: finds the average distance between two points in clusters
• Various distance and similarity functions

Maybe one day I will make some files that show off how to go about using these systems

• Add examples usage files

Feel free to send all comments, questions, or concerns to contact@pepperon.ai