Maximally informative dimensions

The technique of maximizing mutual information between the spike train and the stimulus is described in [1,2].

License

The following terms and conditions apply to use of the Programs. Downloading the programs indicates your acceptance of these terms and conditions.

These programs are provided for non-commercial research use only. The programs will not be distributed further to third parties for any purpose and will not be decompiled, disassembled or otherwise reverse engineered. The Salk Institute makes no representation that the use of the programs will not infringe any patent or other proprietary right.

THE PROGRAMS ARE PROVIDED "AS IS" WITHOUT WARRANTY OF MERCANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE OR ANY OTHER WARRANTY, EXPRESS OR IMPLIED. IN NO EVENT SHALL THE SALK INSTITUTE OR DR. SHARPEE AND DR. KOUH BE LIABLE FOR ANY DIRECT OR CONSEQUENTIAL DAMAGES RESULTING FROM USE OF THE PROGRAMS. THE USER BEARS THE ENTIRE RISK FOR USE OF THE PROGRAMS.

ver 2 (Jan 2010) for finding multiple relevant dimensions

Please download the program code mi2d_2010_01_20.tar.gz (Approx. 1 Mb) and instructions readme.txt .

The code relies on mxml package (mxml-2.6.tar.gz) to input parameters, such as stimulus dimensionality, number of temporal delays, etc.

We also provide a demo, which includes a stimulus set and a model spike train for a cell with two spatiotemporal receptive fields: demo_2010_01_20.tar.gz.

Finally, you may also want to download a set of matlab functions for plotting the results: matlab_2020_01_20.tar.gz.

Older versions: ver 1.4 (June 2009)

Download the codes (Approx. 250 MB).
Untar it (tar -xzvf ).
Read readme.txt for information on how to run from the command line or from within Matlab.
Note: Current version only supports finding ONE maximally informative dimensionat a time. The codes for finding more than 1 dimension simultaneously is provided above.
Extra source codes (from Numerical Recipes in C, 1992): You may need these two functions (ran2.c and nrutil.c), if you need to re-compile the MID codes. You may download these functions only if you own a Numerical Recipes License.

older versions

Ver 1.1 and readme.txt.

References

[1] T. Sharpee, N. Rust, and W. Bialek. "Maximally informative dimensions: Analyzing neural responses to natural signals." in NIPS: 261-268, 2003.
[2] T. Sharpee, N. Rust, and W. Bialek. "Analyzing neural responses to natural signals: Maximally informative dimensions." in Neural Comp.: 16, 223-250, 2004.
[3] T. Sharpee, "Comparion of information and variance optimization strategies for characterizing neural feature selectivity" in Statistics in Medicine.: 26, 4009-4031, 2007.
[4] M. Kouh and T. Sharpee, "Estimating linear-nonlinear models using Renyi divergences", Network: Computationa in Neural Systems.: 20(2):49-68, 2009.

Last modified: Feb 2010.