Elementary network description for efficient link between neuronal models and neuromorphic systems

A method of configuring a neuronal network, the method comprising:

• parsing a specification of a system exhibiting aspects of a neuronal network that includes a plurality of independently executable threads;
• mapping individual ones of the plurality of independently executable threads onto a neuronal network comprising a plurality of doublets, a given doublet being connected to a presynaptic unit and a postsynaptic unit, wherein the given doublet is configured to modify a memory of the postsynaptic unit responsive to events received from the presynaptic unit;
• and
• defining operational characteristics of the plurality of doublets and their corresponding presynaptic and postsynaptic units, wherein
• defining the operational characteristics includes:
• for the given doublet configuring a doublet event rule that determines how the given doublet is configured to update the memory of the postsynaptic unit, wherein execution of doublet event rules is order-independent;
• and
• for a given unit configuring a unit update rule that controls the response of the given unit to memory updates initiated by a doublet, wherein execution of the unit update rules is order-independent;
• and
• wherein mapping individual ones of the plurality of independently executable threads onto a neuronal network comprises:
• providing one or more target libraries that describe units and doublets as constructs in an abstracted description language that is independent of target characteristics;
• and
• mapping individual ones of the plurality of independently executable threads onto the abstracted units and doublets to obtain an abstracted description of the neuronal network.

A system for configuring a neuronal network, the system comprising:

• one or more processors configured to execute computer program modules, the computer program modules comprising:
• a system clock module configured to determine a sequence of time intervals;
• an elements module configured to provide a plurality of elements consisting of units, doublets and triplets, individual units having a memory updated according to a unit update rule of the unit, wherein a doublet event rule configured for a given doublet causes the memory of a postsynaptic unit to be updated in response to an event received from a presynaptic unit, and wherein a given triplet is configured to access the memory of two units and to update the memory of at least one of the two units in accordance with a corresponding triplet update rule;
• and
• a mapper module configured to parse a specification of a neuronal network, and maps individual ones of a plurality of independently executable threads of the neuronal network onto the units, doublets and triplets,
• wherein:
• the mapper module configures the unit update rules, the doublet event rules and the triplet update rules,
• the execution of unit update rules is order-independent and at least partially executed once in every time interval,
• for those doublets that respond to an event during a particular time interval, execution of the doublet event rules corresponding to those doublets is order independent,
• execution of triplet update rules is order-independent and at least partially executed once in every time interval,
• updates to the memory of the units are made by atomic addition;
• and
• for the given doublet, the execution of the corresponding doublet event rule is triggered in response to a change in the memory of the presynaptic unit.