private static void VerifyGeneLength(GeneSequence parent, ICollection<char> childGenes)
{
if (childGenes.Count != parent.Genes.Length)
{
throw new ArgumentException("result is different length from parent");
}
}
public IEnumerator Log(int[] type, bool?wonGame)
{
trialsMasterCount++;
while (reportLockout)
{
yield return(null);
}
reportLockout = true;
GeneSequence ret = population.Find((GeneSequence gs) => gs.Compare((GeneSequence)type) && gs.returned < gs.sent);
//If the returned game was null, that means it did not happen properly, lower this sent amount by one.
if (wonGame == null)
{
ret.sent--;
}
else
{
//The sequence may have been discarded. (This may no longer be necessary since we wait for trials to complete)
if (ret != null)
{
if ((bool)wonGame)
{
ret.successes++;
}
ret.returned++;
//Debug.Log("Log: " + ret.ToString());
}
}
reportLockout = false;
}
private static void ExtractBeginningMinimizers(List <Minimizer> minimizers, GeneSequence sequence, int w, int k)
{
string minimal = sequence.Body.Substring(0, k);
int posMinimal = 0;
if (minimizers.Count == 0 || posMinimal > minimizers[minimizers.Count - 1].Position)
{
minimizers.Add(new Minimizer(minimal, sequence, posMinimal));
}
// since the first kmer is certainly added (it is the only kmer in a size 1 window),
// we can start at position 1, rather than 0,
// and since k has to be at least 1, we can set windowEnd to 2 for beginning
for (int windowEnd = 2; windowEnd < Math.Min(w, sequence.Body.Length - k + 2); windowEnd++)
{
// iterate through all substrings (k-mers) inside a window
for (int i = posMinimal + 1; i < windowEnd; i++)
{
string kmer = sequence.Body.Substring(i, k);
if (String.Compare(kmer, minimal) < 0)
{
minimal = kmer;
posMinimal = i;
}
}
if (minimizers.Count == 0 || posMinimal > minimizers[minimizers.Count - 1].Position)
{
minimizers.Add(new Minimizer(minimal, sequence, posMinimal));
}
}
}
private static void ExtractInteriorMinimizers(List <Minimizer> minimizers, GeneSequence sequence, int w, int k)
{
string minimal = sequence.Body.Substring(0, k);
int posMinimal = 0;
// moving a window, so last iteration has just enough space to fill w kmers
for (int windowStart = 0; windowStart < (sequence.Body.Length + 2 - w - k); windowStart++)
{
if (posMinimal < windowStart)
{
minimal = null;
}
// iterate through all substrings (k-mers) inside a window
for (int j = windowStart; j < windowStart + w; j++)
{
string kmer = sequence.Body.Substring(j, k);
if (minimal == null || String.Compare(kmer, minimal) < 0)
{
minimal = kmer;
posMinimal = j;
}
}
if (minimizers.Count == 0 || posMinimal > minimizers[minimizers.Count - 1].Position)
{
minimizers.Add(new Minimizer(minimal, sequence, posMinimal));
}
}
}
private static void ExtractEndingMinimizers(List <Minimizer> minimizers, GeneSequence sequence, int w, int k)
{
int l = sequence.Body.Length;
string minimal = null;
int posMinimal = l - k;
for (int windowStart = Math.Max(l - k - w, 0) + 1; windowStart <= l - k; windowStart++)
{
if (posMinimal < windowStart)
{
minimal = null;
}
// iterate through all substrings (k-mers) inside a window
for (int i = windowStart; i <= l - k; i++)
{
string kmer = sequence.Body.Substring(i, k);
if (minimal == null || String.Compare(kmer, minimal) < 0)
{
minimal = kmer;
posMinimal = i;
}
}
if (minimizers.Count == 0 || posMinimal > minimizers[minimizers.Count - 1].Position)
{
minimizers.Add(new Minimizer(minimal, sequence, posMinimal));
}
}
}
public bool Compare(GeneSequence other)
{
for (int i = 0; i < genes.Length; i++)
{
if (other[i] != genes[i])
{
return(false);
}
}
return(true);
}
/// <summary>
/// Static function that extracts minimizers from a given sequence
/// </summary>
/// <param name="sequence">Sequence from which to extract minimizers</param>
/// <param name="w">Number of k-mers in one batch (window) of minimizer extraction</param>
/// <param name="k">Minimizer size (character length of substrings (k-mers))</param>
/// <returns>List of extracted minimizers</returns>
public static List <Minimizer> Extract(GeneSequence sequence, int w, int k)
{
if (w < 0 || k < 0 || sequence == null || sequence.Body.Length < k)
{
throw new ArgumentException("Given arguments not valid");
}
List <Minimizer> minimizers = new List <Minimizer>();
ExtractBeginningMinimizers(minimizers, sequence, w, k);
ExtractInteriorMinimizers(minimizers, sequence, w, k);
ExtractEndingMinimizers(minimizers, sequence, w, k);
return(minimizers);
}
private void SetupNodeList(GeneSequence <NodeGene> nodeGenes)
{
var seenBefore = new List <int>();
foreach (var node in nodeGenes)
{
if (!seenBefore.Contains(node.nodeID))
{
seenBefore.Add(node.nodeID);
}
else
{
Console.WriteLine();
foreach (var n in nodeGenes)
{
Console.Write("nID:{0} ", n.nodeID);
}
Console.WriteLine();
throw new Exception("network same nodeID in network!");
}
}
int representedState = 0;
foreach (NodeGene gene in nodeGenes)
{
var nodeID = gene.nodeID;
switch (gene.type)
{
case NodeType.Sensor:
inputNodes.Add(nodeID, new InputNetworkNode(nodeID));
break;
case NodeType.Hidden:
var hnode = new InternalNetworkNode(nodeID, ((InternalNodeGene)gene).Function) as INetworkNode;
awaitingNotificationsNodes[nodeID] = hnode;
hiddenNodes.Add(nodeID, hnode);
break;
case NodeType.Output:
var onode = new OutputNetworkNode(nodeID, representedState++, ((InternalNodeGene)gene).Function) as INetworkNode;
outputNodes.Add(nodeID, onode);
break;
}
}
}
public GeneSequence Generate(IList<GeneSequence> parents, int numberOfGenesToUse, Func<char> getRandomGene, int numberOfGenesInUnitOfMeaning, decimal slidingMutationRate, Func<int, int> getRandomInt, int freezeGenesUpTo)
{
var parent = parents[getRandomInt(parents.Count)];
bool useHint = getRandomInt(2) == 0 &&
parent.Fitness != null &&
parent.Fitness.UnitOfMeaningIndexHint != null &&
parent.Fitness.UnitOfMeaningIndexHint.Value >= freezeGenesUpTo;
int pointA = useHint
? parent.Fitness.UnitOfMeaningIndexHint.Value * numberOfGenesInUnitOfMeaning
: getRandomInt(numberOfGenesToUse - freezeGenesUpTo) + freezeGenesUpTo;
int pointB = getRandomInt(numberOfGenesToUse - freezeGenesUpTo) + freezeGenesUpTo;
if (pointA == pointB)
{
return parent.Clone();
}
var childGenes = parent.Genes.ToArray();
IChildGenerationStrategy type = this;
if (numberOfGenesInUnitOfMeaning == 1 ||
numberOfGenesToUse - freezeGenesUpTo == numberOfGenesInUnitOfMeaning ||
getRandomInt(2) == 0)
{
SwapTwoGenes(childGenes, pointA, pointB);
type = new SwapMidUnitOfMeaning();
}
else
{
pointA /= numberOfGenesInUnitOfMeaning;
pointB /= numberOfGenesInUnitOfMeaning;
SwapTwoUnitsOfMeaning(childGenes, pointA, pointB, numberOfGenesInUnitOfMeaning);
}
VerifyGeneLength(parent, childGenes);
var child = new GeneSequence(childGenes, type);
return child;
}
public IEnumerator GetSpread(System.Action <int[]> callback)
{
while (getLockout)
{
yield return(null);
}
getLockout = true;
if (population == null)
{
population = new List <GeneSequence>();
//Generate initial spread
while (population.Count < initPopSize * 2)
{
//Randomly generate one
int[] spread = new int[geneSize];
int[] defaultSpread = (int[])GameManager.DefaultPointSpread;
//Fill with random values
for (int i = 0; i < spread.Length; i++)
{
spread[i] = Mathf.Max(0, defaultSpread[i] + Random.Range(-fluctationRange - 1, fluctationRange + 1));
}
//Check if it is the same as one already in there, if so, ditch it.
bool isDuplicate = false;
foreach (GeneSequence sequence in population)
{
if (isDuplicate == false)
{
isDuplicate = sequence.Compare((GeneSequence)spread);
}
}
if (isDuplicate == false)
{
//Debug override always use default point spread
//population.Add((GeneSequence)defaultSpread);
population.Add((GeneSequence)spread);
}
}
timeStamp = Time.time;
originTimeStamp = Time.time;
Debug.Log("Simulation start");
for (int i = 0; i < population.Count; i++)
{
Debug.Log("Init: " + population[i].ToString());
}
}
GeneSequence ret = population.Find((GeneSequence gs) => gs.sent < trials);
if (ret == null)
{
Debug.LogWarning("All trials sent." + (Time.time - timeStamp));
//Wait for all trials to complete
foreach (GeneSequence item in population)
{
while (item.returned < item.sent)
{
yield return(null);
}
}
Debug.LogWarning("Trial Complete." + (Time.time - timeStamp));
timeStamp = Time.time;
//Sort the old population by how well they performed
population.Sort((GeneSequence a, GeneSequence b) => {
float aVal = (float)a.successes / (float)a.returned;
float bVal = (float)b.successes / (float)b.returned;
if (aVal < bVal)
{
return(-1);
}
else if (aVal == bVal)
{
return(0);
}
else
{
return(1);
}
});
//Trim the worst ones
while (population.Count > initPopSize)
{
population.RemoveAt(0);
}
float lowest = population[0].successes / (float)population[0].returned;
float highest = population[population.Count - 1].successes / (float)population[population.Count - 1].returned;
population.Reverse();
for (int i = 0; i < population.Count; i++)
{
Debug.Log("C Result: " + population[i].ToString());
}
int diffCount = 0;
for (int i = 0; i < population.Count; i++)
{
for (int k = 0; k < population.Count; k++)
{
if (population[i].Compare(population[k]) == false)
{
diffCount++;
}
}
}
if (diffCount < convergenceThreshold * initPopSize)
{
//We have converged
Debug.LogError("Convergence " + diffCount + ". Total Time: " + (Time.time - originTimeStamp));
Debug.Break();
Application.Quit();
}
else
{
//Randomize the population
int n = population.Count;
while (n > 1)
{
n--;
int k = Random.Range(0, n + 1);
GeneSequence value = population[k];
population[k] = population[n];
population[n] = value;
}
List <GeneSequence> children = new List <GeneSequence>();
//Go in pairs and breed new children.
for (int i = 0; i < population.Count; i += 2)
{
if (i + 1 < population.Count)
{
GeneSequence a = population[i];
GeneSequence b = population[i + 1];
int[] childArray = new int[a.genes.Length];
//Get a random value from each parent
for (int k = 0; k < childArray.Length; k++)
{
childArray[k] = (Random.value < 0.5f) ? a.genes[k] : b.genes[k];
}
//Mutate
if (Random.value < mutationRate)
{
Debug.Log("Mutation");
//Get a random position
int pos = Random.Range(0, childArray.Length);
//Shift it up or down
if (Random.value < 0.5f)
{
childArray[pos]++;
}
else
{
childArray[pos]--;
}
}
children.Add((GeneSequence)childArray);
}
}
//Add t he children to the population
foreach (GeneSequence child in children)
{
population.Add(child);
}
}
for (int i = 0; i < population.Count; i++)
{
population[i].sent = 0;
population[i].returned = 0;
population[i].successes = 0;
}
//select the first thing
ret = population[0];
}
ret.sent++;
//Select an sequence with not enough trials.
callback(ret.genes);
getLockout = false;
}
public GeneSequence Generate(IList<GeneSequence> parents, int numberOfGenesToUse, Func<char> getRandomGene, int numberOfGenesInUnitOfMeaning, decimal slidingMutationRate, Func<int, int> getRandomInt, int freezeGenesUpTo)
{
var indexes = Enumerable.Range(0, parents.Count)
.Shuffle().Take(2).ToArray();
int i1 = indexes.First();
int i2 = indexes.Last();
int sliceIndex = getRandomInt(numberOfGenesToUse - freezeGenesUpTo) + freezeGenesUpTo;
var parentA = parents[i1];
IChildGenerationStrategy type = this;
if (numberOfGenesInUnitOfMeaning > 1 &&
numberOfGenesToUse - freezeGenesUpTo != numberOfGenesInUnitOfMeaning &&
getRandomInt(2) == 1)
{
bool useHint = getRandomInt(2) == 0 &&
parentA.Fitness != null &&
parentA.Fitness.UnitOfMeaningIndexHint != null &&
parentA.Fitness.UnitOfMeaningIndexHint.Value >= freezeGenesUpTo;
sliceIndex = useHint
? parentA.Fitness.UnitOfMeaningIndexHint.Value * numberOfGenesInUnitOfMeaning
: sliceIndex - sliceIndex % numberOfGenesInUnitOfMeaning;
}
else
{
type = new SpliceMidUnitOfMeaning();
}
if (sliceIndex == 0)
{
return parentA.Clone();
}
var parentB = parents[i2];
var childGenes = parentA.Genes.Take(sliceIndex).Concat(parentB.Genes.Skip(sliceIndex)).ToArray();
VerifyGeneLength(parentA, childGenes);
var child = new GeneSequence(childGenes, type);
return child;
}