Device for the autonomous bootstrapping of useful information

A self-learning, artificial neural network based-discovery device, comprising:

• a first artificial neural network portion capable of generating patterns;
• means for subjecting said first artificial neural network to varying degrees of perturbation;
• means for implementing at least one of said generated patterns of said first self-learning artificial neural network as a concept;
• means of evaluating an effect of said concept;
• means of triggering reinforcement learning within said first self-training artificial neural network of one of said generated patterns leading to a positive effect.

A self-learning, artificial neural network based-discovery device, comprising:

• an array of self-learning artificial neural networks, each of said artificial neural networks capable of generating patterns in response to varying degrees of perturbation to it;
• means for subjecting at least one of said self-learning artificial neural networks to varying degrees of perturbation;
• wherein at least a first, perturbed, self-learning artificial neural network is associated with at least a second of said self-learning artificial neural networks such that said second of said self-learning artificial neural networks receives generated patterns from said first perturbed artificial neural network;
• means for implementing said generated patterns of at least said first perturbed self-learning artificial neural network as a concept;
• means of evaluating an effect of said concept;
• means of triggering reinforcement learning within at least said first perturbed, self-training artificial neural networks upon said generated patterns leading to a positive effect;
• means of triggering reinforcement learning in at least said second self-learning artificial neural network so as to incorporate within said second network a relationship between at least one of said generated patterns of said first, perturbed, self-learning artificial neural network and said effect of said generated pattern.

A self-learning, artificial neural network-based discovery device, specific to path planning, comprising:

• a first self-learning artificial neural network capable of producing generated patterns in response to perturbations applied to said first network and wherein said generated patterns represent potential paths through an environment;
• means for subjecting said first self-learning artificial neural network to varying degrees of perturbation;
• means for calculating a cost associated with said potential paths;
• means for providing environmental inputs to said means for calculating a cost, wherein said means for calculating uses said environmental inputs to generate said cost;
• and
• wherein said means for calculating a cost determines a minimal cost path through said environment from said potential paths generated by said perturbed first neural network.

An artificial neural network-based discovery device, comprising:

• a first artificial neural network capable of generating patterns in response to perturbations applied to said first network and wherein said generated patterns represent potential plans of action;
• means for subjecting said first artificial neural network to varying degrees of perturbation;
• means for calculating a cost associated with said potential plans of action;
• means for providing environmental inputs to said means for calculating a cost, wherein said means for calculating uses said environmental inputs to generate said cost;
• and
• wherein said means for calculating a cost determines a minimal cost plan of action from said potential plans of action generated by said perturbed first neural network.

A self-learning, artificial neural network-based discovery device, comprising:

• a first, self-learning artificial neural network capable of generating patterns in response to perturbations applied to said first network;
• means for subjecting said first self-learning artificial neural network to varying degrees of perturbation;
• means of determining a value of said generated patterns;
• and
• means of triggering reinforcement learning within said first self-training artificial neural network of at least one of said generated patterns having a desired value.

A self-learning, artificial neural network-based device for object detection, comprising:

• means for providing sensor perspectives of an environment;
• means to move an attention window, wherein each position of said attention window represents at least one of said sensor perspectives of said environment;
• and
• a first self-learning artificial neural network whose input patterns consist of content of said sensor perspectives and operative to generate a recognition metric indicating recognition of said content of said sensor perspectives.

An artificial neural network-based device, comprising:

• a first, auto-associative, artificial neural network trained upon patterns having the commonality of belonging to a particular group or genre;
• a comparator portion, said comparator portion being operative to compare patterns supplied to said first network with patterns generated by said first network in order to produce a reconstruction error associated with said first artificial neural network;
• and
• wherein said reconstruction error represents a recognition metric of patterns belonging to said group or genre.

An artificial neural-network based navigational field generation system, comprising:

• a first artificial neural network;
• means for providing environmental sensor inputs to said first artificial neural network;
• and
• wherein said first network is operative to generate a navigational field comprising field values at coordinates within said environment.