//Initialize the first training cycle result
public TrainingCycleResult(ItemType[] supportedTypes)
{
Aborted = false;
Networks = new Dictionary <ItemType, ItemNetwork>();
foreach (ItemType type in supportedTypes)
{
Networks[type] = new ItemNetwork(new ParsedItem[0], new KnownAffix[0], new KnownAffix[0]);
}
}
private TrainingCycleResult PerformCycle(Snapshot snapshot)
{
//Cancel right away when a more recent snapshot has been provided
if (snapshot.Id < LastReceivedSnapshotId)
{
Console.WriteLine($"({snapshot.Id}) Evaluation of snapshot was aborted since a more recent snapshot is present.");
return(TrainingCycleResult.ABORTED);
}
Console.WriteLine($"({snapshot.Id}) Evaluating snapshot...");
//Initialize a new dictionary to hold the loaded items
Dictionary <ItemType, List <ParsedItem> > loadedItems = new Dictionary <ItemType, List <ParsedItem> >();
foreach (ItemType type in SupportedItemTypes)
{
loadedItems[type] = new List <ParsedItem>();
}
foreach (JsonPOEStash s in snapshot)
{
foreach (JsonPOEItem i in s.Items)
{
//Skip unsupported items
if (!SupportedItemTypes.Contains(i.ItemType))
{
continue;
}
//No uniques for now
if (i.FrameType != FrameType.Magic && i.FrameType != FrameType.Rare)
{
continue;
}
//Ignore non-chaos priced items for now
if (i.Price.IsEmpty() || i.Price.CurrencyType == CurrencyType.Unknown)
{
continue;
}
//Convert the value if needed
float value;
if (i.Price.CurrencyType != CurrencyType.ChaosOrb)
{
value = conversionTable.ConvertTo(i.Price, CurrencyType.ChaosOrb).Value;
}
else
{
value = i.Price.Value;
}
ParsedItem b = new ParsedItem(i.Corrupted, i.ImplicitMods, i.ExplicitMods)
{
CalibrationPrice = value
};
loadedItems[i.ItemType].Add(b);
}
}
foreach (ItemType type in SupportedItemTypes)
{
Console.WriteLine($"({snapshot.Id}) Number of {type} items loaded: {loadedItems[type].Count}. This cycle there were {loadedItems[type].Count - lastCycleResult.Networks[type].CalibrationItemsCount} new items added.");
}
Dictionary <ItemType, ItemNetwork> networks = new Dictionary <ItemType, ItemNetwork>();
foreach (ItemType type in SupportedItemTypes)
{
var knownImplicits = new Dictionary <string, KnownAffix>();
var knownExplicits = new Dictionary <string, KnownAffix>();
foreach (ParsedItem i in loadedItems[type])
{
foreach (ParsedAffix a in i.ParsedImplicitMods)
{
KnownAffix knownAffix;
if (!knownImplicits.TryGetValue(a.AffixCategory, out knownAffix))
{
knownAffix = new KnownAffix(a.AffixCategory, a.Value, a.Value);
knownImplicits.Add(knownAffix.AffixCategory, knownAffix);
}
knownAffix.UpdateWith(a);
}
foreach (ParsedAffix a in i.ParsedExplicitMods)
{
KnownAffix knownAffix;
if (!knownExplicits.TryGetValue(a.AffixCategory, out knownAffix))
{
knownAffix = new KnownAffix(a.AffixCategory, a.Value, a.Value);
knownExplicits.Add(knownAffix.AffixCategory, knownAffix);
}
knownAffix.UpdateWith(a);
}
}
ItemNetwork network = new ItemNetwork(
loadedItems[type].ToArray(),
knownImplicits.Values.ToArray(),
knownExplicits.Values.ToArray());
Console.WriteLine($"({snapshot.Id}) Training the {type} network...");
network.LearnFromItems();
networks[type] = network;
}
Console.WriteLine($"({snapshot.Id}) Training of snapshot {snapshot.Id} was done.");
//Print the info of this run
foreach (ItemType type in SupportedItemTypes)
{
Console.WriteLine($"({snapshot.Id}) Accuracy of {type} network with {networks[type].CalibrationItemsCount} items: {(int)networks[type].DetermineAccuracy()} %");
}
Console.WriteLine(string.Empty);
return(new TrainingCycleResult(networks));
}
