"Spike timing dependent plasticity to make robot navigation more intelligent."

some implementation is like if synapse i is excitatory/inhibitory ... then ... but actually LTP and LTD can occur at the same synapse so that this can self-stabilize a network

N-methyl-D-aspartate

Hebbian STDP has a particular sensitivity to causality: if a presynaptic neuron
contributes to the firing of the postsynaptic neuron, the plasticity mechanism
will strengthen the synapse, and thus the presynaptic neuron will become more
effective in causing the postsynaptic neuron to fire. This mechanism can determine
a network to associate a stable output to a particular input. But let us imagine
that the causal relationships are reinforced only when this leads to something
good (for example, if the agent controlled by the neural network receives a positive
reward),while the causal relationships that lead to failure are weakened, to avoid
erroneous behavior.


Self-organization-and-dynamics-reduction-in-recurrent-networks sitmulus-presentation-and-learning