< Back to Article

Efficient "Shotgun" Inference of Neural Connectivity from Highly Sub-sampled Activity Data

Fig 1

Observation scheme examples.

In each scheme (the different rows) we observe two out of ten neurons in each time bin: (A,B) Fixed subset (C,D) Serial (E,F) Fully Random, (G,H) Random Blocks, and (I,J) double serial. Left (A,C,E,G,I): A sample of the observations O demonstrating the scanning method (a zero-one matrix, Eq 3). Right (B,D,F,H,J): empirical frequency of observed neuron pairs , with saturated colors to accentuate differences between methods (all values above 0.05 are shown in yellow). In the “fixed” scheme, some neurons are never observed. In the “serial” scheme some neuronal pairs are never observed. In all other schemes, all neuronal pairs are observed, so we can estimate the empirical moments using Eqs (16)–(17) and infer connectivity. In the two bottom schemes, observations are collected in persistent blocks, so neuron pairs which are close to the diagonal are observed more often.

doi: https://doi.org/10.1371/journal.pcbi.1004464.g001