Intelligent control with hierarchical stacked neural networks
First Claim
1. A method of processing information, comprising:
- receiving a message;
processing the message with a trained artificial hierarchical stacked neural network implemented with at least one automated processor, comprising a plurality of hierarchical layers, each respective hierarchical layer being trained according to at least one training criteria, the at least one training criteria differing for respective ones of the plurality of hierarchical layers, and comprising at least one set of outputs which together represent a non-arbitrary organization of actions based on an architecture of the respective hierarchical layer and the prior training of the respective hierarchical layer;
the message being received as an input by a respective first hierarchical layer of the trained artificial hierarchical stacked neural network;
generating a search query for at least one database external to the hierarchical stacked neural network, based on the message, the search query being generated within the hierarchical stacked neural network as a neural network layer output from a respective second hierarchical layer of the trained artificial hierarchical stacked neural network at a respectively higher hierarchical level than the respective first hierarchical level which receives the message;
communicating the search query through an automated communication channel, wherein the at least one database is configured to receive the search query from the automated communication channel and produce a database response based on an index of a plurality of database records corresponding to the search query;
receiving the database response from the at least one database through the automated communication channel, as a neural network input into a respective third hierarchical layer of the trained artificial hierarchical stacked neural network;
generating an output as a non-arbitrary organization of actions from a respective fourth hierarchical layer of the trained artificial hierarchical stacked neural network, selectively in dependence on the database response.
0 Assignments
0 Petitions
Accused Products
Abstract
A method of processing information is provided. The method involves receiving a message; processing the message with a trained artificial neural network based processor, having at least one set of outputs which represent information in a non-arbitrary organization of actions based on an architecture of the artificial neural network based processor and the training; representing as a noise vector at least one data pattern in the message which is incompletely represented in the non-arbitrary organization of actions; analyzing the noise vector distinctly from the trained artificial neural network; searching at least one database; and generating an output in dependence on said analyzing and said searching.
1173 Citations
20 Claims
-
1. A method of processing information, comprising:
-
receiving a message; processing the message with a trained artificial hierarchical stacked neural network implemented with at least one automated processor, comprising a plurality of hierarchical layers, each respective hierarchical layer being trained according to at least one training criteria, the at least one training criteria differing for respective ones of the plurality of hierarchical layers, and comprising at least one set of outputs which together represent a non-arbitrary organization of actions based on an architecture of the respective hierarchical layer and the prior training of the respective hierarchical layer; the message being received as an input by a respective first hierarchical layer of the trained artificial hierarchical stacked neural network; generating a search query for at least one database external to the hierarchical stacked neural network, based on the message, the search query being generated within the hierarchical stacked neural network as a neural network layer output from a respective second hierarchical layer of the trained artificial hierarchical stacked neural network at a respectively higher hierarchical level than the respective first hierarchical level which receives the message; communicating the search query through an automated communication channel, wherein the at least one database is configured to receive the search query from the automated communication channel and produce a database response based on an index of a plurality of database records corresponding to the search query; receiving the database response from the at least one database through the automated communication channel, as a neural network input into a respective third hierarchical layer of the trained artificial hierarchical stacked neural network; generating an output as a non-arbitrary organization of actions from a respective fourth hierarchical layer of the trained artificial hierarchical stacked neural network, selectively in dependence on the database response. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9)
-
-
10. A system for processing information, comprising:
-
an input port receiving configured to receive at least one message; and a trained artificial hierarchical stacked neural network, implemented with at least one automated processor, comprising a plurality of hierarchical layers, each respective hierarchical layer being trained according to at least one training criteria, the at least one training criteria differing for respective ones of the plurality of hierarchical layers, and comprising at least one set of outputs which together represent a non-arbitrary organization of actions based on an architecture of the respective hierarchical layer and a prior training of the respective hierarchical layer, the trained artificial hierarchical stacked neural network being configured to; receive the at least one message as an input to a respective first hierarchical layer of the hierarchical stacked neural network; generate a search query for at least one database external to the hierarchical stacked neural network, based on the message, the search query being generated within the hierarchical stacked neural network as a neural network output at a respective second hierarchical layer of the hierarchical stacked neural network, at a respectively higher hierarchical level of the hierarchical stacked neural network than the respective first hierarchical level which receives the message as an input; communicate the search query through an automated communication channel, wherein the at least one database is configured to receive the search query from the automated communication channel and produce a database response based on an index of a plurality of database records corresponding to the search query; receive the database response from the at least one database through the automated communication channel, as a neural network input into a respective third hierarchical layer of the hierarchical stacked neural network; and generate an output as a non-arbitrary organization of actions from at least one respective fourth layer of the trained artificial hierarchical stacked neural network, selectively in dependence on the received database response. - View Dependent Claims (11, 12, 13, 14, 15, 16, 17)
-
-
18. A method of automatically carrying out a sensory-motor task, comprising:
-
receiving an input comprising environmental information generated by an automated environmental sensor; interpreting abstract information represented in the environmental information within a hierarchical stacked artificial neural network comprising at least three separately trainable hierarchical layers, each of the at least three separately trainable hierarchical layers being trained according to at least one set of training criteria, the at least one set of training criteria differing for respective ones of the at least three separately trainable hierarchical layers; conducting a database search of a database external to the hierarchical stacked artificial neural network comprising a plurality of records having content defined independently of the hierarchical stacked artificial neural network, based on a search query generated as an output by the hierarchical stacked artificial neural network; and generating control information relating to a manner of carrying out the sensory-motor task by an automated controller based at least on a received result of the database search. - View Dependent Claims (19, 20)
-
Specification