Temporally dynamic artificial neural networks
An apparatus, article and method containing an artificial neural network that, after training, produces new trainable nodes such that input data representative of a first event and input data representative of a second event both activate a subset of the new trainable nodes. The artificial neural network can generate an output that is influenced by the input data of both events. In various embodiments, the new trainable nodes are sequentially produced and show decreasing trainability over time such that, at a particular point in time, newer produced nodes are more trainable than earlier produced nodes. The artificial neural network can be included in various embodiments of methods, apparatus and articles for use in predicting or profiling events.
1. A method of profiling a user-defined event, comprising:
providing a trained artificial neural network that produces new trainable nodes by creating a first subset of new trainable nodes based on data representative of a first event, the first event occurring at a first time instant, and by creating a second subset of new trainable nodes based on data representative of a second event, the second event occurring at a second time instant different from the first time instant, such that input data representative of the first event and input data representative of the second event both activate a same set of the new trainable nodes if the first event and the second event are within a threshold time of each other;
generating from the artificial neural network an output that is influenced by the input data of both events; and
profiling a user-defined event based on the output of the artificial neural network.
2. The method of claim 1 , wherein the input data representative of the first event and the input data representative of the second event are each representative of a person or a physical object.
3. The method of claim 1 , wherein the user-defined event relates to information processing, or to an activity having observable dynamics over long time scales.
4. The method of claim 3 , wherein the user-defined event relates to: a) demand for or usage of area codes, zip codes, telephones, cell phones, internet services, power infrastructure, or traffic patterns and flow; b) dynamics of financial systems, stock markets, commodities markets, or options markets; c) profiling systems for use in homeland security, advertising, or medical records applications; d) data compression or encryption; e) sports forecasting; f) weather forecasting; g) genomic analysis; h) brain networks or brain network modeling; i) robotics; or j) military applications.
5. The method of claim 1 , wherein the trainable nodes are continually produced.
6. The method of claim 1 , wherein the trainable nodes are trainable for a predetermined time period.
7. The method of claim 1 , wherein, after being produced, the new trainable nodes show decreasing trainability over time.
8. The method of claim 7 , wherein the new trainable nodes are sequentially produced such that, at a particular point in time, newer produced nodes are more trainable than earlier produced nodes.
9. A computer system for profiling an event, comprising:
system memory;
a trained artificial neural network stored on the system memory and configured to produce new trainable nodes by creating a first subset of new trainable nodes based on data representative of a first event, the first event occurring at a first time instant, and by creating a second subset of new trainable nodes based on data representative of a second event, the second event occurring at a second time instant different from the first time instant, such that input data representative of the first event and input data representative of the second event both activate a same set of the new trainable nodes if the first event and the second event are within a threshold time of each other, and wherein an output of the artificial neural network is influenced by the input data of both events; and
one or more processors configured to implement the trained artificial neural network;
wherein the computer system profiles a user-defined event based on outputs of the artificial neural network.
10. The computer system of claim 9 , wherein the user-defined event relates to information processing, or to an activity having observable dynamics over long time scales.
11. The computer system of claim 10 , wherein the user-defined event relates to: a) demand for or usage of area codes, zip codes, telephones, cell phones, internet services, power infrastructure, or traffic patterns and flow; b) dynamics of financial systems, stock markets, commodities markets, or options markets; c) profiling systems for use in homeland security, advertising, or medical records applications; d) data compression or encryption; e) sports forecasting; f) weather forecasting; g) genomic analysis; h) brain networks or brain network modeling; i) robotics; or j) military applications.
12. The computer system of claim 9 , wherein the trainable nodes are continually produced.
13. The computer system of claim 9 , wherein the trainable nodes are trainable for a predetermined time period.
14. The computer system of claim 9 , wherein, after being produced, the new trainable nodes show decreasing trainability over time.
15. The computer system of claim 14 , wherein the new trainable nodes are sequentially produced such that, at a particular point in time, newer produced nodes are more trainable than earlier produced nodes.
16. A non-transitory computer-readable storage medium with computer executable instructions stored thereon, the instructions for performing a method of profiling a user-defined event, the method comprising:
providing a trained artificial neural network that produces new trainable nodes by creating a first subset of new trainable nodes based on data representative of a first event, the first event occurring at a first time instant, and by creating a second subset of new trainable nodes based on data representative of a second event, the second event occurring at a second time instant different from the first time instant, such that input data representative of the first event and input data representative of the second event both activate a same set of the new trainable nodes if the first event and the second event are within a threshold time of each other;
generating from the artificial neural network an output that is influenced by the input data of both events; and
profiling a user-defined event based on the output of the artificial neural network.
17. The computer-readable storage medium of claim 16 , wherein the user-defined event relates to information processing, or to an activity having observable dynamics over long time scales.
18. The computer-readable storage medium of claim 17 , wherein the user-defined event relates to: a) demand for or usage of area codes, zip codes, telephones, cell phones, internet services, power infrastructure, or traffic patterns and flow; b) dynamics of financial systems, stock markets, commodities markets, or options markets; c) profiling systems for use in homeland security, advertising, or medical records applications; d) data compression or encryption; e) sports forecasting; f) weather forecasting; g) genomic analysis; h) brain networks or brain network modeling; i) robotics; or j) military applications.
19. The computer-readable storage medium of claim 16 , wherein the trainable nodes are continually produced.
20. The computer-readable storage medium of claim 16 , wherein the trainable nodes are trainable for a predetermined time period.
21. The computer-readable storage medium of claim 16 , wherein, after being produced, the new trainable nodes show decreasing trainability over time.
22. The computer-readable storage medium of claim 21 , wherein the new trainable nodes are sequentially produced such that, at a particular point in time, newer produced nodes are more trainable than earlier produced nodes.