SciPy Probability Distributions Visualizer

Visualize probability distributions from SciPy using Plotly-Dash. I made this to review and better understand various probability distributions, and to serve as a handy tool to use alongside scikit-learn's RandomizedSearchCV when tuning machine learning models.

• Probability Distributions
• SciPy Stats Reference
• Plotly Dash

Demo

Freely hosted here, availability not guaranteed.

---

Contents

• Code Notes
  • Overview
  • For Me
• Installation
  • WSGI

Code Notes

Overview

• app.py
  • load distributions from SciPy, setup dash application, handle callbacks
• distributions.py
  • functions to
    • update info text and parameter inputs on page with distribution choice
    • generate histogram and violin plots from given distribution and parameters
      • five thousand points generated: <distribution>.rvs(<shape_parameters>,<loc>,<scale>,5000)

For Me

• Python
  • class inspection to extract distributions and their information from SciPy
• SciPy
  • experience with probability distributions, code, documentation
• Dash, Plotly
  • graphs with custom hover data, colors, titles
    • only use Go figures, so plotly-express and pandas are not required packages
  • advanced callback usage. | works, imperfectly
    • distribution selection from dropdown determines parameters (and info text) which determine graphs
    • graphs updated by sliders or text inputs
    • sliders and text boxes update if their corresponding partner is changed
      • circular callback warning, causes graphs to update twice each time
  • simple single page application, easy deployment
  • minimal CSS styling, no CSS/JSS toolkits like Bootstrap
    • could use something to display LaTeX, SciPy docs utilize sphinx which uses MathJax?

---

Installation

• Requirements File lists all packages frozen in my development environment
• install dash and SciPy in a Python environment to run the applcation

The code can be run from a virtual environment on your local machine. Download the code to a new directory, your-directory

Only the python files .py and assets (your-directory/assets) directory are necessary to run the application.

# Adapt for your operating system

cd your-directory
python -m venv venv
venv\scripts\activate
python -m pip install --upgrade pip

# install packages via requirements.txt
pip install -r requirements.txt 
# OR install packages separately
pip install dash scipy

# ensure server settings are ok before running
python -m app

The application will be available on the default port, 8050, at http://127.0.0.1:8050/. Change DEBUG settings or adapt deployment for WSGI by modying the run command.

WSGI

• comment out run command
• import app from app.py into your WSGI of choice
  • I think PythonAnywhere uses uWSGI with NGINX