SAModel

class spikeometric.models.SAModel[source]

Bases: spikeometric.models.base_model.BaseModel

The Synaptic Activation model (SAModel) is a base model for models that use the synaptic activation as the state of the network and has an update rule for based on previous synaptic activation and spikes.

In addition to the input, non_linearity and emit_spikes methods, SAModels must implement the update_activation method.

update_activation(spikes, activation)[source]

The update rule for the synaptic activation.

simulate(data: torch_geometric.data.data.Data, n_steps: int, verbose: bool = True, equilibration_steps: int = 100, store_as_dtype: torch.dtype = torch.int32)[source]

Simulates the network for n_steps time steps given the connectivity. Returns the state of the network at each time step.

Parameters

  • data (torch_geometric.data.Data) – The data containing the connectivity.

  • n_steps (int) – The number of time steps to simulate

  • verbose (bool) – If True, a progress bar is shown

  • equilibration (int) – The number of time steps to simulate before starting to record the state of the network.

  • store_as_dtype (torch.dtype) – The dtype to store the state of the network

Returns

x – The state of the network at each time step. The state is a binary tensor where 1 means that the neuron is active.

Return type

torch.Tensor[n_neurons, n_steps]

tune(data, firing_rate, tunable_parameters='all', lr=0.1, n_steps=100, n_epochs=100, verbose=True)[source]

Tunes the model parameters to match a desired firing rate.

Parameters

  • data (torch_geometric.data.Data) – The training data containing the connectivity.

  • firing_rate (torch.Tensor) – The target firing rate of the network

  • tunable_parameters (list or str) – The list of parameters to tune, can be “all”, “stimulus”, “model” or a list of parameter names

  • lr (float) – The learning rate

  • n_steps (int) – The number of time steps to simulate for each epoch

  • n_epochs (int) – The number of epochs

  • verbose (bool) – If True, a progress bar is shown

equilibrate(edge_index: torch.Tensor, W: torch.Tensor, initial_state: torch.Tensor, n_steps=100, store_as_dtype: torch.dtype = torch.int32) → torch.Tensor[source]

Equilibrate the network to a given connectivity matrix.

Parameters

  • edge_index (torch.Tensor) – The connectivity of the network

  • W (torch.Tensor) – The connectivity filter

  • initial_state (torch.Tensor) – The initial state of the network

  • n_steps (int) – The number of time steps to equilibrate for

  • store_as_dtype (torch.dtype) – The dtype to store the state of the network as

Returns

x – The state of the network at each time step

Return type

torch.Tensor