ConceptNetwork

Related: ConceptCircuit, SoftwareModule

Combination of SoftwareNeurons and SoftwareSynapses that might perform some interesting/useful function.

Every neuron and synapse has a specific state in the network depending on its ConceptNeuronModel and ConceptSynapseModel (which defines what will happen next).

Information from one network can be included in other networks. This means that if you create some useful functionality in a small number of neurons, replicating this functionality many times is a fairly simple prospect.

// Neurons in the brain don’t have specific addresses and are not accessible in a specific order. But for SoftwareModule they have been made addressable by SoftwareNeuronIndex or SoftwareNeuronLabel.

File

A network-state can be saved to files and restored for future computation. (SoftwareNetworkFile)

List of Current Networks (v1.0)

• 3DSim—demonstrates the 3D simulator under development.
• BabyTalk—learns to speak by trial-and-error learning. Where SpeechTest works with words, BabyTalk works with phonemes.
• BasicNeurons.xml—illustrates the simplest neuron models.
  • 450-neurons
• CameraTest—inputs the camera on your computer to neuron values.
• Imagination—shows how Sallie's internal mental model can be used for imagination.
• Maze—shows navigation of a maze the way a 3-year-old might. As Sallie explores the maze, she remembers landmarks, the decisions she made, and the results which were achieved. These are the necessary components of reinforcement learning.
• NeuralGraph—demonstrates how a mathematical graph can be implemented in neurons controlled by a module.
• ObjectMotion—demonstrates how Sallie can move objects in her environment.
• Sallie—is an end-to-end AGI model using hearing, vision, and knowledge to learn the meanings of a few words.
• SimVision—shows Sallie's simulated environment and vision.
• SpeechTest—uses speech recognition and synthesis.