After releasing the first version of the toolbox, I have been receiving emails from users all over the world. It was exciting and rewarding to see so much interest building around such a simple tool.
From the feedback, however, I realized that maybe the software was actually too simple! The lack of structure and documentation were sometimes misleading users into applying the software incorrectly or even when they didn't really needed the intricacies of information-theory for their analyses.
For this reason I have decided to discontinue the first version of the toolbox, and to start working on a new much enriched tool. This will be featuring:
- A clear guide with everything you need for successfully estimating information from samples: From the the sampling-bias problem, to issues regarding binning and number of trials.
- Completely automated binning;
- Fully-integrated Matlab-style documentation and support;
- Matlab-style functions, input-structure, names and so on. The goal is to achieve the level of integration with Matlab in which one can simply change the name of a function and be able to compare the information results with those of corrcoef;
- Arbitrary dimensionality for all variables (not just what is currently called the "response") involved in the calculations. This will also allow loosening the link between the toolbox and neuroscience;
The goal is to develop a generic and comprehensive package for "numerical" information theory.You can find a tentative Table of Contents of the new toolbox here (this will be regularly updated).
It will probably take a "bit" to get the whole thing together, however I am sure that the results will be worth the waiting! Meanwhile, feel free to contact me for questions or suggestions [address: my-name dot my-surname at the-company-that-provides-the-world's-most-famous-internet-search dot com].
I look forward to meet you again on these pages,
P.S.: yes, the website is under construction ...