Sometimes it’s better to actually see the influences of the parameters of distributions. This holds true especially for more complex ones. Hence, I build an app that lets you play around with some of the most ubiquitous distributions. We are starting with a bare minimum of some well-known distributions and will successively add more. Please tell me in the comments below, which distribution you would like to see implemented next.
Probability distributions are ubiquitous in the machine learning world and, e.g. at the core of whole fields such as Bayesian learning. So before we start, let us quickly recapture what distributions are and what they are good for in a machine learning context. Here are some descriptions from the net:
In probability theory and statistics, a probability distribution is a mathematical function that provides the probabilities of occurrence of different possible outcomes in an experiment.Wikipedia
So basically, it is a list of values that a random variable can take on and its corresponding probability. In machine learning, probability densities are often used to connect data and desired outcome. If the density has parameters, e.g. such as mean and variance in the Gaussian distribution, the overall goal is often to find the best set of parameters to fit the underlying data.
That should be enough and without further ado, have some fun playing with distributions, you guys.
29. Nov 2019
- added Gamma distributions
28. Nov 2019
- starting with Gaussian and Laplace distribution
About the app
The app is written in Python and using the streamlit framework for the interactive components. Matplotlib is used for plotting the graphics and descriptions are downloaded directly from wikipedia using the corresponding public API. The main feature of the app is its ability to implement other distributons fast without having to fiddle with lists and views. It is a very simple and efficient setup using decorators. All of the resources and source codes are available in the /streamlit-app folder for 123AI.de members.
If you like it, leave a comment and tell your friends. Also, if you can’t help yourself, head over to my steady page and consider supporting my work which would be awesome!