# All the model and the analyses were implemented in MATLAB.


-getActiveUnits.m
---This function return the active units indexes according to the output of a layer.

-getLifeSparse.m
---This function return the lifetime sparseness of each unit according to the output of a layer.

-getPSI.m
---This function return the PSI matrix according to the distributions of every unit's responses to every phoneme.

-getSTRF.m
---This function return the strfs of each unit in the first 3 S layers according to the model parameters

-getSTRFFeatures
---This function return the 4 STRF features according to a STRF

-main.m
---This is the script file for SHMAX setup, training and inference. The corresponding parameters are determined in the "setParameters.m".

-sample3dImages.m
---This function return random 3D patches from the data

-sampleImages.m
---This function return random 2D patches from the data

-SHMAX_2dS.m
---This function return the bases and do the inference according to the model parameters, which is used for 2D inputs.
---The training algorithm is implemented by the "SPAMS"*.

-SHMAX_3dS.m
---This function return the bases and do the inference according to the model parameters, which is used for 3D inputs.

-SHMAX_C.m
---This function do the max pooling according to the model parameters.

-setParameters.m
---This is a script file that determines the parameters used for build the deep SHMAX model.

* SPAMS: http://spams-devel.gforge.inria.fr/