Future research direction based on this project is threefold.
First, the algorithm could be generalized to any order Markov process, at the
expense of
requiring larger amount of data.
Second, we would like to derive the theoretical error bound for the estimate
in terms of basic information-theoretic quantities such as mutual information
and entropy.
Third, it may not be necessary to decode from spike trains, as there is some
evidence, at least for certain neural systems, that there is a
significant amount of information in local field potential (LFP). Throwing
away the LFP might be a tremendous wastage of useful data. An important
question is the level of redundancy between the LFP and the spikes, and
the role of LFP in neural systems, which some researchers hypothesize
lie in spike synchronization.
Here is a list of people involved in this project:
- Shiyan Cao
- Michael Epstein
- Zoran Nenadic
A useful introductory material on the theory of point processes can be
found
here
To read about the basics of the new decoding algorithm click
here