ANT - Adapting Neural Topology

This implementation is a Neuralevolution (NE) type approach.

• Add the package to your app.

Use ANT inside your XCode project:

• Create a ANT network.

import ANT


let xor = ANT(1024)
xor.initializeEntities(structure: EntityStructure(inputs: 2, outputs: 1, inputActivation: .sigmoid, outputActivation: .sigmoid))

Example:

let inputs: [[Float32]] = [[0.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 1.0]]
let expOutput: [[Float32]] = [[0.0], [1.0], [1.0], [0.0]]

var highestScore: Float32 = 0

while true {
    
    for e in 0..<xor.entities.count {
        
        // Do test for entity
        var scoreTotal: Float32 = 0.0
        
        for i in 0..<inputs.count { // for each training value
            
            let result = xor.learn(entityIndex: e, input: inputs[i])
            
            for o in 0..<result.count {
                scoreTotal += abs(expOutput[i][o] - result[o])
            }
        }
        xor.appendCurrentEntityScore(entityIndex: e, pow(4 - scoreTotal, 2))

        if xor.entities[e].score > highestScore {
            xor.king = xor.entities[e]
            highestScore = xor.entities[e].score
        }
        
    }
    
    if highestScore >= (16 * 0.98) { break }

    xor.nextIteration()     // Next iteration of entities.
}

// Iterate test throught the king entity.
for i in 0..<inputs.count { print(xor.passThroughKing(input: inputs[i])) }

for c in xor.king!.network.connections {
    print("from: \(c.from), to: \(c.to), weight: \(c.weight) ")
}
for layer in xor.king!.network.layers {
    print("ID: \(layer.id), Nodes: \(layer.currentNodeIndex - 1), Activation: \(layer.activationFunction)")
}