Octave Etard ([email protected])
Matlab routines for fitting linear forward / backward models / CCA.
Requires Matlab R2019b or newer (tested on R2019b & R2020a/b).
This package consists of Matlab routines for fitting linear forward / backward models (deconvolution). It implements ridge-regularised linear models, with options to use different L2 penalties (e.g. curvature). This implementation strives for computational and memory efficiency to allow for more flexibility e.g. when fitting large models with high sampling rate, or to form generic cross-validated models faster.
This is achieved by using Fourier methods to compute cross-covariance / cross-correlation matrices directly, without ever forming the underlying matrices.
It incorporates high-level "wrapper" functions that should make fitting models straightforward, with users needing only to specify functions to read-in the input data, and contains examples illustrating how to use the code.
Add the functions folder to your path. The code is structured as a Matlab package contained in functions/+LM to reduce the risk of shadowing any of the user's own functions. Functions can be called by using the LM. prefix (e.g. out = LM.someFunction(x,y,z)). Alternatively the required functions can also be imported (e.g. import LM.someFunction; out = someFunction(x,y,z)).
Note a large number of calls of the type LM.someFunction in a script can lead to decreased performance. This can be fixed by encapsulating the calls in a function or by importing the relevant functions.
The example folder contains detailed examples illustrating how to use this package based on synthetic data. These include:
example_spikes_deconvolution.m: deconvolution example that showcase the use of the low-level routines.example_forward.m&example_backward.m: simple fitting of forward/backward models, using the wrapper functions so that one does not have to deal with the lower level functions mentioned above.
For more examples, have a look at LMpackage_examples, which demonstrates the use of this package to fit forward / backward models with cross-validation (subject-specific and generic (population) models) on real EEG datasets.