This package contains the software tools and the hardware specifications used to develop a bidirectional brain-machine interface featuring a neuromorphic hardware decoder.

This software is to fit FitzHugh-Nagumo (FHN) model of state dependence of neural responses was developed in paper:

[1]  H. Safaai, R. Neves, O. Eschenko, N. K. Logothetis, Stefano Panzeri, A dynamical model of  the effect of Locus Coeruleus firing on single-trial cortical state dynamics and sensory, PNAS 2015, published ahead of print September 28, 2015, doi:10.1073/pnas.1516539112

Any person downloading this software accepts to acknowledge this study by citing it in every publication or report arising from the use of this material. 
The software can be downloaded as rar file from this link.
In the software, the file Test.m fits to cortical S1 multiunit activity (MUA) the FHN dynamical model with Locus Coeruleous (LC) neuromodulatory inputs to a 1.5 sec stretch of spontaneous MUA activity.
The function test.m calls the following functions contained in the package:
-       FHN_Auto.m: Function which gets the model parameters and modulatory inputs and integrates the dynamics prediction using the auto-dynamical equations defined in [1].
-       FHN_Self.m: Similar to FHN_Auto but this functions computes the self-dynamical equations defined in SAfaai et al (PNAS 2015).
-       FHN_ep_Auto.m: Computes the cost functions of the model fit error and the MUA for auto-dynamical model.
-       FHN_ep_Auto.m: Similar to FHN_ep_Auto.m but for self-dynamical model.
-       FHN_OPT.m: Optimizes the fit error and estimates the self and auto dynamical model parameters. 
-      test.m: The function calls an example dataset DATA.m that contains Matlab data with the following structure used to run a test example of the codes:

    DATA.MUA: MUA activity
    DATA.Time: Time stams
    DATA.Input1: Input 1 to the dynamical model
    DATA.Input2: Input 2 to the dynamical model
    DATA.PARAMS: Estimated model parameters of the example trial.

Here we show the link to the software for space-by-time separable non-negative matrix factorization that was developed by Delis and colleagues and published in the following article:

Delis I., Panzeri S., Pozzo T., Berret B. (2014) A unifying model of concurrent spatial and temporal modularity in muscle activity, J Neurophysiol 111: 675–693

This software can be found at:
http://hebergement.u-psud.fr/berret/software/sNM3F.zip

Here we report the code developed and used in the paper:

Cavallari S, Panzeri S and Mazzoni A (2014) Comparison of the dynamics of neural interactions between current-based and conductance-based integrate-and-fire recurrent networks. Front. Neural Circuits 8:12. doi: 10.3389/fncir.2014.00012

The code is made available in the ModelDB sharing repository:
http://senselab.med.yale.edu/ModelDB/ShowModel.asp?model=152539
with accession number 152539

This software computes mutual information between stimuli and neural response both after discounting the state variability and without considering the state variability. This software also provides a simulation of Poisson neurons that responds to stimuli and that has an additive state dependent component. This simulation may be useful to users to try out the software and test some ideas. These methods developed and used in papers:

 H. Safaai, R. Neves, O. Eschenko, N. K. Logothetis, Stefano Panzeri, A dynamical model of  the effect of Locus Coeruleus firing on single-trial cortical state dynamics and sensory, PNAS 2015, ; published ahead of print September 28, 2015, doi:10.1073/pnas.1516539112

C. Kayser, C. Wilson, H. Safaai, S. Sakata, S. Panzeri, Rhythmic Auditory Cortex Activity at Multiple Timescales Shapes Stimulus–Response Gain and Background Firing, The Journal of Neuroscience, 20 May 2015, 35(20): 7750-7762; doi: 10.1523/JNEUROSCI.0268-15.2015

Any person downloading this software accepts to acknowledge this study by citing it in every publication or report arising from the use of this material. 

The folder contains the following set of functions to compute the information of the discounted trial-to-trial variability.

STATE_INFORMATION.m: The function computes the following three information values

Information between the stimulus and response I(S,R)

Information between the stimulus and discounted variability response I(S;R’) where the trial-to-trial variability V is subtracted from the response as R’=R-V.

The joint information between the stimulus and the response and state I(S;R,theta)

State_Variability_Simulator.m: The functions simulates responses of a poisson process with the trial-to-trial variability which is a linearly related to the phase of the ongoing activity at the moment of stimulus onset. The function first simulates the data, then estimates the trial-to-trial variability by fitting the data to the linearly additive state model and then apply the fit to a cross-validated test set. This function uses the functions lint.m and fit_lintr.m functions to fit the linear additive state model to the data.

Main.m: This is the main file which can be used to either simulate the trial-to-trial variability or use the DATA.mat file as the input of the STATE_INFORMATION.m function. The output will be a bar plot of different information values.

The data that underlie the main text and supplementary figures of this paper, as well as the code for the transfer entropy calculation, can be found at https://dx.doi.org/10.6084/m9.figshare.1460872