Connectome noise¶
This package implements connectome noise models.
-
class
connectome.noise.combined_noise.RemoveAddNoiseAndSubsample(**kwargs)¶ Inputs: network, fraction_remove_and_add, subsampling_fraction
Subsample a fraction of the network and shuffle edges in the remaining portion.
- Parameters
network (
Network) – The network.fraction_remove_and_add (float in [0, 1]) – Fraction of edges to be removed and randomly reinserted again. The sign of an edge is automatically converted to match the type of the new presynaptic neuron.
subsampling_fraction (float in [0, 1]) – Fraction of the network to keep. subsampling_fraction=1 implies the full network.
-
construct_graph()¶ Construct the computational graph here.
-
class
connectome.noise.barrelcutnoise.BarrelCutNoise(**kwargs)¶ Inputs: f, p, network
Simulate to cut a barrel into two pieces. Also simulate network noise. For spatially embedded networks, the embedding is taken into account.
- Parameters
f – float Fraction of the network to be removed.
p – float Noise applied to the network after cutting. The noise type is
RemoveAndAddEdgesNoise. The parameter p is its parameter fraction_remove_and_add.
-
edge_removal_probability()¶ Uses the empirical distribution \(emp\) from 1:
Number synapses per connection
Probability
1
0
2
2/11
3
5 /11
4
2/11
5
2/11
The expected value is 3.36.
Assumes that synapses are uniformly distributed throughout the barrel. Therefore, probability of having a synpase on the “same” side is \((1-f)\).
The fraction of edges additionally lost due to the cut is
\[p_{cut}(f) = \sum_{n=0}^5 Binom(0 |n, f)emp(n).\]- Parameters
f (float) – Fraction of the barrel which is still left.
- 1
Feldmeyer, Dirk, Veronica Egger, Joachim Lübke, and Bert Sakmann. “Reliable Synaptic Connections between Pairs of Excitatory Layer 4 Neurones within a Single ‘barrel’ of Developing Rat Somatosensory Cortex.” The Journal of Physiology 521, no. 1 (1999): 169–90. doi:10.1111/j.1469-7793.1999.00169.x.
-
class
connectome.noise.edgeshufflingnoise.EdgeShufflingNoise(**kwargs)¶ Inputs: network, fraction_draws_over_nr_synapses
Shuffle edges randomly.
This is an abstract class implementing a template method pattern.
- Parameters
network (
Network) – The network to be shuffled.fraction_draws_over_nr_synapses (int) – The parameter fraction_draws_over_nr_synapses multiplied with the number of synapses (edges) in the network is the to be rounded number of reshuffling attempts.
-
abstract
_shuffle_subpopulation(submatrix, diagonal)¶ Template method for subclasses.
- Parameters
submatrix (view on array) – The part to be reshuffled.
diagonal (bool) – True if it contains the diagonal part of the complete array.
-
class
connectome.noise.inandoutdegreepreservingnoise.InAndOutDegreePreservingNoise(**kwargs)¶ Inputs: network, fraction_draws_over_nr_synapses
Shuffle edges in a way such that in and out degrees are preserved.
This is a subclass of
EdgeShufflingNoise.The shuffling is achieved through a Markov-Chain-Monte-Carlo method. An ergodic Markov chain on the space of graphs is constructed. It is then sampled from this Markov chain. The sample is thus an unbiased sample from all graphs with the given in and out degree distribution.
-
class
connectome.noise.outdegreepreservingnoise.OutDegreePreservingNoise(**kwargs)¶ Inputs: network, fraction_draws_over_nr_synapses
Shuffle a network such that all out degrees are preserved.
This is a subclass of
EdgeShufflingNoise.
-
class
connectome.noise.removeandaddedgesnoise.RemoveAndAddEdgesNoise(**kwargs)¶ Inputs: network, fraction_remove_and_add
Remove and add edges uniformly.
- Parameters
network (
Network) – The network to which the noise is applied.fraction_remove_and_add (float in [0, 1]) – The fraction of edges to be removed and randomly reinserted.
