[ANN]: DiscretePredictors.jl - A Package for online prediction of discrete symbols


#1

Hello everyone,

I’m very happy to announce a julia package (my first one :slight_smile: ) for online discrete symbol prediction DiscretePredictors.jl.

DiscretePredictors is a collection of algorithms for online discrete sequence prediction. It provides a common API to various discrete sequence prediction algorithms. My goals with the package are:

  • Simple Interface
  • Easily Extendable
  • Reproducible Research

Documentation is available at: https://v-i-s-h.github.io/DiscretePredictors.jl/stable/

Kindly share your views about the package.

With Regards
vish


#2

What is online? And discrete sequence prediction? A simple high-level summary would be nice. I have a Master of Statistics but have never come across these concepts.


#3

My guess would be that online refers to the fact that data is fed continuously. In certain frameworks (eg neural networks) one distinguishes between batch learning and online learning.
Discrete: I would guess that there is a finite number of possible symbols.


#4

Hi,
Online-discrete here means that you are seeing a sequence of symbols and you want to predict what will be the next symbol based on your history of observations. For a more detailed explanation, checkout http://www.eecs.wsu.edu/~cook/pubs/is05.pdf

This package includes some algorithms which can be used in such scenarios.


#5

That’s what prediction is, generally :smile:

I agree with @bernhard, and suspect that online in this context means that previous observations are summarized by a sufficient statistic that is simpler than the space of all possible histories (similarly to OnlineStats.jl; this is a common usage of the term), and discrete means that the set of possible observations is countable (and potentially finite).