Hard-coded baseline correction

[katduecker]

The different resting membrane potentials in the apical_tuft and soma of the L5 pyramidal cells cause a non-zero dipole at rest. Additionally, the compartments are not initialized at their resting membrane potentials, which causes an initial "drift" in the dipole.

The baseline correction is hard-coded in dipole.py https://github.com/jonescompneurolab/hnn-core/blob/master/hnn_core/dipole.py#L611

This baseline correction will not work for new models.

I understand we don't want to add a baseline correction to each simulation based on a buy-in interval, as this would add a lot of time and compute to each simulation.

As a current work-around, I have run my new model for 5 seconds (could do 10) without any inputs, stored the L2 and L5 dipoles and time vector in a .json file, which is then subtracted during a baseline correction.

def _baseline_renormalize_ca(self):
        """Baseline correction based on calcium model without drives"""

        hnn_core_root = op.dirname(hnn_core.__file__)

        # load the baseline dipole
        with open(op.join(hnn_core_root, 'param', 'bsl_dipole_ca.json'), 'r') as f:
            bsl_dpl = json.load(f)

        # subtract baseline dipole from the current dipole
        for i,t in enumerate(self.times):
            bsl_idx = np.argmin(np.abs(bsl_dpl['times'] - t))
            self.data['L2'][i] -= bsl_dpl['L2'][bsl_idx]
            self.data['L5'][i] -= bsl_dpl['L5'][bsl_idx]
        
        self.data['agg'] = self.data['L2'] + self.data['L5']

I have further added an option to simulate_dipole to select a baseline correction method, which defaults to jones (the original correction)

def simulate_dipole(net, tstop, dt=0.025, n_trials=None, record_vsec=False,
                    record_isec=False, record_ca=False, postproc=False, bsl_cor='jones')

Again, this is just a work-around that works for my current purposes, and I'm happy to discuss ideas for a more permanent solution.