Build your own Scikit-NeuroMSI model

You can implement your own model by importing the SKNMSIMethodABC method from the core module and creating a class calling such method.

Such class must have four attributes to define the model’s identity and input/output modalities:

_model_name: Name of the model.
_model_type: Classification of the model (e.g. MLE, Bayesian, Neural, etc).
_run_input: Modality specific parameters computed by the model.
_run_output: Modality specific results delivered by the model.

Furthermore, the class must have six managed attributes (properties) defined in the init method:

mode0: Name of sensory modality 0.
mode1: Name of sensory modality 1.
position_range: Range of positions encoded by the model.
position_res: Resolution of positional encoding.
time_range: Range of times encoded by the model.
time_res: Resolution of time encoding.

The model may include more than two sensory modalities, as long as they are properly defined in the class.

Finally, the class must have two methods to specify the model execution:

set_random: Sets random seed (if required).
run: Executes the model

Here a template of a Scikit-NeuroMSI model class:

[1]:

from skneuromsi.core import SKNMSIMethodABC


class MyBayesianModel(SKNMSIMethodABC):
    _model_name = "MyBayesianModel"
    _model_type = "Bayesian"
    _output_mode = "multi"

    _run_input = [
        {"target": "auditory_position", "template": "${mode0}_position"},
        {"target": "visual_position", "template": "${mode1}_position"},
    ]

    _run_output = [
        {"target": "auditory", "template": "${mode0}"},
        {"target": "visual", "template": "${mode1}"},
    ]

    def __init__(
        self,
        *,
        mode0="auditory",
        mode1="visual",
        position_range=(0, 29),
        position_res=1,
        time_range=(1, 2),
        time_res=1,
    ):
        self._mode0 = mode0
        self._mode1 = mode1
        self._position_range = position_range
        self._position_res = float(position_res)
        self._time_range = time_range
        self._time_res = float(time_res)

    # PROPERTY ================================================================

    @property
    def mode0(self):
        return self._mode0

    @property
    def mode1(self):
        return self._mode1

    @property
    def time_range(self):
        return self._time_range

    @property
    def time_res(self):
        return self._time_res

    @property
    def position_range(self):
        return self._position_range

    @property
    def position_res(self):
        return self._position_res

    # MODEL RUN ================================================================

    def set_random():
        pass

    def run(auditory_position, visual_position): ...

For more details about model building, please refer to the API documentation.