Back to results
US 12,124,954 B1Granted
Intelligent control with hierarchical stacked neural networks
OwnerMichael Lamport Commonsinventor
Life of this patent
Estimating the term…
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; and analyzing the noise vector distinctly from the trained artificial neural network.
Claims20
An artificial neural network system, comprising:
- an input port configured to receive input data;
- a first neural network, comprising at least one automated processor and a memory configured to store coefficients defining a first input layer, at least one first hidden layer, and a first output layer of the first neural network, trained with training data to produce the stored coefficients connecting the first input layer to the at least one first hidden layer, and connecting the at least one first hidden layer to the first output layer, to produce:
- a first nonarbitrary organization of actions at the first output layer having an organization selectively dependent on the training of the first neural network with the training data, and being responsive to a first subset of information present in the input data and unresponsive to a residual second subset of information present in the input data, selectively dependent on a feedback signal, and
- a first arbitrary action not being organized selectively dependent on the training data, comprising a first noise vector output with respect to the first nonarbitrary organization of actions and which comprises the residual second subset of information present in the input data;
- a second neural network, comprising the least one automated processor and a memory configured to store data defining a second input layer, at least one second hidden layer, and a second output layer of the second neural network, trained to produce a second nonarbitrary organization of actions and the feedback signal, selectively dependent the first nonarbitrary organization of actions.
2 more elements
An artificial neural network method, comprising:
- receiving environmental sensor data through an input port;
- providing a first neural network, comprising a first input layer, a first hidden layer, and a first output layer, trained with training data to produce a first nonarbitrary organization of actions at the first output layer selectively dependent on environmental sensor data and a feedback signal received at the first input layer;
- providing a second neural network, comprising a second input layer, a second hidden layer, and a second output layer, trained to produce a second nonarbitrary organization of actions and the feedback signal at the second output layer selectively dependent the first nonarbitrary organization of actions received at the second input layer;
- generating the first nonarbitrary organization of actions dependent on the environmental sensor data and the feedback;
- generating the second nonarbitrary organization of actions and the feedback dependent on the first nonarbitrary organization of actions;
- and
- generating a noise vector from at least one of the first neural network and the second neural network comprising a processed arbitrary organization of actions dependent on information from the input layer of the respective neural network which is incompletely represented in the non-arbitrary organization of actions of the respective neural network.
4 more elements
An artificial neural network, comprising:
- an input port configured to receive real time environmental sensor data;
- a stacked neural network, comprising a plurality of hierarchically organized neural networks, each respective neural network comprising an input layer, at least one hidden layer, and an output layer, each respective neural network being trained to produce a nonarbitrary organization of actions organized dependent on the training, which is passed to a hierarchically superior neural network and at least one respective neural network having a feedback signal that passed to a hierarchically inferior neural network;
- and
- at least one respective neural network producing, in addition to the nonarbitrary organization of actions organized dependent on the training, an arbitrary organization of actions organized independent of the training, comprising a noise vector comprising processed information from the input layer of the respective neural network which is incompletely represented in the non-arbitrary organization of actions of the respective neural network.
1 more element
Dependencies are inferred from the claim text.