I have attached a draft paper explaining the math behind frequency coding, a method of lossless data compression I came up with. The algorithm has been tested as I have freely available computer source code for anyone who wants to test it. The paper still needs to be complete and is missing the following:

- Math Cleanup – my terminology needs to be re-worked, its vague and hard to understand 🙂
- Decoding Process – important for paper proof – shown in source code on my website.
- Probability comparison – chart of actual p(x), frequency encoding output, neural network stretch = ln(x / (1 – x))
- Calgary Corpus Comparison – compare to other arithmetic encoders on this standard data set

This algorithm is fairly intuitive once you understand the concept behind it.

Here is the PDF document: