Is it possible to build an associative memory (say Correlation Matrix Memory or CMM) using Leaky Integrate and Fire style spiking neurons that uses the STDP learning algorithm? It should be able to associate a time ordered sequence of spike pulses (forming a neural code) with another, such that on giving the first sequence of spikes as input we get an output sequence of spikes that can be decoded into the associated sequence.

Some ideas:

My previous approach: Hebbian learning, use CMM with rank order codes (weight matrix of 1, alpha, alpha squared etc) and new weight matrix is max of old and the transpose of the two vectors to be associated. Dont know if this is true STDP approach, or if others have tried something like this, or if this is biologically plausible.

Just extending this: a model of associative memory that can be microscoped up to sequence memory level, sequence memory applications level, down to neuron level, spike level, ultimately Calcium and Na ions level.