作者: Kaeser M. Sabrin , Yongbin Wei , Martijn van den Heuvel , Constantine Dovrolis
DOI: 10.1101/600999
关键词: Sensory system 、 Artificial neural network 、 Connectome 、 Network analysis 、 Flow (mathematics) 、 Neuroscience 、 Feed forward 、 Synapse 、 Hourglass 、 Gap junction 、 Computer science
摘要: Abstract We approach the C. elegans connectome as an information processing network that receives input from about 90 sensory neurons, processes through a highly recurrent of 80 interneurons, and it produces coordinated output almost 120 motor neurons control nematode’s muscles. apply recently developed analysis framework (“hourglass effect”) [1], focus on feedforward flow to neurons. The reveals this traverses small core waist”) consists 10-15 interneurons – mostly same set were previously shown constitute “rich-club” [2]. This result is robust methodology separates feedback information. remains when we consider only chemical synapses or combination gap junctions. hourglass organization suggests has some similarities with encoder-decoder artificial neural networks in which vector first compressed low-dimensional latent space encodes data more efficient manner, followed by decoding intermediate-level salient functions are combined different ways compute correlated outputs network.