WhoBPyT New Features Plan

[JohnGriffiths]

We have a new set of models and ideas to be implemented in the library.

They are all variants of linear eigenmode formulations of neural mass models of various kinds.

Making notes here on a plan for implementing these pieces.

This is the order we'll do things in:

1. Linear power spectrum models [ Robinson , JR, conductance-based, etc. ]
2. Linear FC matrix models [ with joint eigenspace stuff ]
3. Spatial eigenmode-based neural mass models
4. EEG/MEG extended spatial inverse idea
5. Time-varying eigenvalue estimation

Zheng to do for Mon 7th April:

For number 1:

• What are the new files, functions, classes that are going to be added
• Write on this thread a high-level summary of what the new changes are going to
• ...before we go ahead with a [ long ] series of PRs

[JohnWangDataAnalyst]

whobpyt/whobpyt
/models/
robinson_fq
robinson_fq.py
class ParamsRobinsonFreq(AbstractParams):
define model parameter use Par
set attributes
class LINEAR_FQ(AbstractNMM)
forward(self, input): from given frequency to power based on closed-loop transfer function

linearfq_2p
linearfq_2p.py
class ParamsLinearFreqs(AbstractParams):
define model parameter use Par
set attributes

            class ROBINSON_FQ(AbstractNMM):
                     forward(self, input): from given frequency to power based on closed-loop transfer function

run/
add modelfitting.py for frquency domain
optimization/
Cost_TS.py
add option for taking log for MSE