Differences

This shows you the differences between two versions of the page.

--- manual:duty_cycle [2014/11/05 15:45] – zenke
+++ manual:duty_cycle [2017/01/24 21:39] (current) – adds note zenke
@@ Line 3: / Line 3: @@
 Once a simulation is triggered by the ''run()'' directive in [[System]] the simulation evolves according to a fixed loop structure.
-First [[System]] will call all the ''evolve()'' methods of all registered instances of [[SpikingGroup]]. During this step all dynamic ODEs of the the neuron and synapse models are updated. Then in a second step it calls all ''propagate()'' methods of instances of [[Connection]] objects. This generally takes care of propagating spikes produced by any [[SpikingGroup]] and delivers them to [[NeuronGroup|NeuronGroups]] where they have their effect on the specified transmitter target (see [[TransmitterType]]). Now all [[Monitor|Monitors]] are called via their ''propagate()'' function. This is mainly to write results to files or do some online monitoring (see [[Checker]]). In a final step the ''evolve()'' function of [[Connection]] objects are called. This is generally to implement continuous weight updates caused for instance by a synaptic decay.
+Auryn's duty cycle can be coarsely divided into three main parts.
+  - **Evolve** the state of the network are evolved by one time step. To that end System calls the ''evolve()'' methods of all registered instances of [[SpikingGroup]] and all [[Connection]] instances. During this step all the smooth dynamics (ODEs) of the the neuron and synapse models are updated. Eg. all integrate-and-fire neurons will smoothly upgrade their internal state.
+  - **Propagate** this step captures the interactions between the different network entities. Mostly, the step is concerned with bringing spikes from point A to point B, where this might entail sending spikes from one node to another in a parallel simulation. To implement propagate, System calls all ''propagate()'' methods of instances of [[Connection]] objects. The nature of spikes can lead to "jumps" in the state of the system, for instance because arriving spikes increase the membrane potential of a target neuron.
+  - **Execute devices**. During this step all [[Monitor|Monitors]] are called via their ''execute()'' function. This is mainly to write results to files or do some on-line monitoring (see [[Checker]]). Note, that in older Auryn versions the ''execute'' methods is also called ''propagate'' for Monitor classes.