================================================================================================== R. V. Florian (2004) "Evolution of alternate object pushing in a simulated embodied agent: Preliminary report." ================================================================================================== "At each time step, only the neurons that receive spikes are updated (hence the event driven nature of the updating of the network). If the updated neurons fire, their spikes are stored in a list. This spike list will be used during the next" "The spikes of the motor neurons were converted to an analog value by a leaky integrator of time constant  = 10ms. The maximum value of the effector activation, 1, corresponds to a firing rate of the motor neuron of 100Hz." "Following Di Paolo (2002b), the analog values of the sensor activations were converted to a spike train using a Poisson process with a firing rate proportional to the activation. The maximum firing rate of the input neurons was set to 100Hz." -------------------------------------------------------------------------------------------------- "Each sensor of the agent, of activation s: 0<=s<=1, drove two input spiking neurons, one being fed with activation s and the other with activation 1-s. Thus, both the activation of the sensor and its reciprocal was fed to the network, and there are 70 input neurons in the network. The reason of this duplication of the sensory signal in the spiking neural network is twofold. First, this allows the network to be active even in the absence of sensory input. For example, if the agent is in a position where nothing activates its sensors (there is no object in its visual range, no tactile contact etc.), there must be however some activity in the neural network, in order for the effectors to be activated and the agent to orientate to stimuli. Second, this mechanism implies that the total input of the network is approximately constant in time (the number of spikes that are fed to the network by the input)." -------------------------------------------------------------------------------------------------- => It was not shown in (Di Paolo 2002b) at least explicitly. Di Paolo, E. A. (2002a) "Evolving spike-timing dependent plasticity forrobot control" EPSRC/BBSRC International Workshop: Biologicallyinspired Robotics, The Legacy of W. Grey Walter, WGWf2002. HP Labs, Available from: http://www.cogs.susx.ac.uk/users/ezequiel/dipaolo wgw2002.ps. => Not available any more -------------------------------------------------------------------------------------------------- Sen Song In the simulations, excitatory synapses are modified based on their pre- and postsynaptic spike timing, whereas inhibitory synapses are held fixed.