Methods and systems for implementing dynamic neural networks

A data processing system comprising:

• a non-transitory computer readable medium storing computer readable instructions and a data structure configured to compute a desired dynamic function, wherein the data structure comprises a plurality of nodes, each node having a node input and a node output, and the plurality of nodes being arranged into a plurality of layers of nodes including at least one input layer and at least one output layer;
• and
• a computer processor operable to execute the computer readable instructions stored on the computer readable medium using the data structure to compute the desired dynamic function;
• wherein the data structure is defined by
• identifying a system-specific synaptic filter H of the computer processor;
• determining the desired dynamic function ƒ to be computed by the plurality of nodes;
• defining at least one node response function for the plurality of nodes, wherein the at least one node response function is defined to receive varying inputs over time at each node input and produce varying outputs over time at each node output;
• associating a plurality of temporal filters with the plurality of nodes, wherein at least one particular temporal filter of the plurality of temporal filters is associated with each node in the plurality of nodes;
• determining a mapping function corresponding to the system-specific synaptic filter H, wherein the mapping function is selected from a plurality of potential mapping functions, wherein the plurality of potential mapping functions include a first set of mapping functions usable to account for discrete time synaptic filters and a second set of mapping functions usable to account for continuous time synaptic filters;
• determining a modified dynamic function ƒ′ based on the desired dynamic function ƒ using the selected mapping function, wherein the modified dynamic function ƒ′ is determined to account for the system-specific synaptic filter H;
• and
• using the at least one node response function and the at least one temporal filter to determine connection weights between the plurality of nodes in the artificial neural network to match the modified dynamic function ƒ′, wherein each connection weight is coupled to at least one of a corresponding node input or a corresponding node output and each connection weight is configured to be multiplied by at least one of the input received by the corresponding node input or the output produced by the corresponding node output whereby the connection weights enable the plurality of nodes to compute the desired dynamic function ƒ based on input received at the at least one input layer.

A computer program product comprising a non-transitory computer readable medium storing computer readable instructions for implementing an artificial neural network for dynamic computing applications on a given computing system, the artificial neural network comprising a plurality of nodes, each node having an input and an output, wherein the computer readable instructions are defined to configure a processor to:

• specify a desired dynamic function ƒ to be computed;
• define at least one node response function for the plurality of nodes, wherein the at least one node response function is defined to receive varying inputs over time at each node input and produce varying outputs over time at each node output;
• associate a plurality of temporal filters with the plurality of nodes, wherein at least one particular temporal filter of the plurality of temporal filters is associated with each node in the plurality of nodes;
• determine a system-specific synaptic filter H of the given computing system, wherein the system-specific synaptic filter H is one of a discrete time filter and a continuous time filter;
• determine a modified dynamic function ƒ′ based on the desired dynamic function ƒ and the system-specific synaptic filter H, wherein the modified dynamic function ƒ′ is determined to account for the system-specific synaptic filter H;
• use the at least one node response function and the plurality of temporal filters to determine connection weights between the plurality of nodes in the artificial neural network to match the modified dynamic function ƒ′, wherein each connection weight is coupled to at least one of a corresponding node input or a corresponding node output and each connection weight is configured to be multiplied by at least one of the input received by the corresponding node input or the output produced by the corresponding node output;
• and
• operate the artificial neural network on the given computing system using the determined connection weights to compute the desired dynamic function ƒ;
• wherein the computer readable instructions enable the processor to account for the system-specific synaptic filter H when the system-specific synaptic filter H is the discrete time filter and when the system-specific synaptic filter H is the continuous time filter.