search

aimacode / aima-java

Java implementation of algorithms from Russell And Norvig's "Artificial Intelligence - A Modern Approach"
MIT License
1.56k stars 796 forks source link

Support for Continuous random Variables #428

Open samagra14 opened 6 years ago

samagra14 commented 6 years ago

Currently, the code repository supports only discrete random variables. The code should be extended to include continuous random variables. Support for some common probability density functions (for eg: Normal distribution, geometric distribution etc. ) must also be added.

belerico commented 5 years ago

I would like to contribute. Can you briefly tell me where to start? Sincerely Federico Belotti

samagra14 commented 5 years ago

Hi @belerico As a first step, I would like to suggest you go through the probability module from the code repo. When we work with any particular Bayesian Network, the random variables at various nodes can be both discrete as well as continuous. At present, we have assumed that the random variables represented by the nodes are discrete and hence we represent their conditional probability distributions using a categorical table. The final aim should be to enable the use of continuous random variables in the Bayesian Network. We can begin with the support of a Gaussian Random Variable and add other distributions later on. Feel free to ping if you need any help.