Tag-based apparatus and methods for neural networks
Apparatus and methods for high-level neuromorphic network description (HLND) using tags. The framework may be used to define nodes types, define node-to-node connection types, instantiate node instances for different node types, and/or generate instances of connection types between these nodes. The HLND format may be used to define nodes types, define node-to-node connection types, instantiate node instances for different node types, dynamically identify and/or select network subsets using tags, and/or generate instances of one or more connections between these nodes using such subsets. To facilitate the HLND operation and disambiguation, individual elements of the network (e.g., nodes, extensions, connections, I/O ports) may be assigned at least one unique tag. The tags may be used to identify and/or refer to respective network elements. The HLND kernel may comprises an interface to Elementary Network Description.
1. A computer realized method of implementing a neural network comprising a plurality of elements, the method comprising:
generating said plurality of elements;
identifying a subset of said plurality of elements using a tag; and
assigning said tag to each element of said subset;
wherein said generating precedes said assigning said tag;
wherein the tag comprises a unique identifier configured to identify said each element; and
wherein said assigning said tag is configured to enable generation of a new network element comprising at least a portion of elements of said subset.
2. The method of claim 1 , wherein said each element is selected at random from said plurality of elements.
3. The method of claim 1 , wherein said each element of said subset comprises a unit.
4. The method of claim 1 , wherein said tag comprises sting identifier.
5. The method of claim 1 , wherein said tag comprises an alphanumeric identifier.
6. The method of claim 5 , wherein said alphanumeric identifier is adapted to identify a spatial coordinate of respective element of said subset.
7. The method of claim 5 , wherein;
said subset comprises a plurality of nodes; and
said alphanumeric identifier comprises an identifier of at least one node of said plurality of nodes.
8. The method of claim 5 , wherein said tag is adapted to enable identification of said subset.
9. The method of claim 1 , wherein said new network element comprises a connection.
10. The method of claim 9 , wherein said connection comprises one of:
(i) a synapse, or
(ii) a junction.
11. A method of dynamic partitioning of a computerized neural network comprising a plurality of elements, the method comprising:
identifying a subset of elements of said network using a tag;
wherein said identifying and said assigning cooperate to enable selection of said each element of said subset using a single selection operation; and
wherein said identifying said subset is based at least in part on executing a Boolean expression comprises a keyword selected from the group consisting of AND, NOT, and OR.
12. The method of claim 11 , further comprising assigning said tag to said subset.
13. The method of claim 11 , further comprising assigning said tag to each element of said subset.
14. The method of claim 11 , wherein said subset comprises a plurality of nodes of said plurality of elements.
15. The method of claim 11 , further comprising:
identifying one other subset of elements of said network using one other tag; and
enabling a plurality of connections between at least a portion of elements within of said subset and elements of said one other subset.
16. The method of claim 15 , further comprising assigning said one other tag to each element of said one other subset.
17. The method of claim 15 , wherein each connection of said plurality of connections comprises one of synapse and junction.
18. The method of claim 15 , wherein each confection of said plurality of connection is enabled based at least in part on said tag and said one other tag.
19. The method of claim 15 , wherein at least a portion of elements within said one other subset being different from elements of said subset.
20. A processing apparatus comprising a nonvolatile storage medium configured to store a plurality of instructions, which, when executed, effect dynamic partitioning of a neural network according to a method, the method comprising:
identifying a subset of elements of said neural network;
executing, by the processing apparatus, a mathematical expression configured to identify each element of said subset; and
assigning a tag to said each element of said subset of elements, said tag comprising an identifier configured to identify said each element;
wherein said assigning said tag is configured to enable generation of a new network element comprising said subset of elements; and
wherein said mathematical expression comprises a Boolean operation.
21. The apparatus of claim 20 , wherein the method is implemented using an Application Specific Integrated Circuit (ASIC) using ASIC instruction set.
22. The apparatus of claim 20 , wherein said each element of said subset is selected using a random selection operation.
23. The apparatus of claim 20 , wherein the method further comprises assigning said tag to said new network element.
24. The apparatus of claim 20 , wherein said assigning said tag to said subset is configured to enable representation of said network as a directed graph.
25. The apparatus of claim 20 , wherein the method further comprises assigning a second tag to said subset, said second tag being distinct from said tag.