static void Main(string[] args)
{
if (args.Length != 6)
{
LogHelper.Log("Invalid arguments!");
LogHelper.Log("Example) Core.exe [Corpus Path] [Alpha] [Beta] [Topic Count] [Inference Iteration Step] [Model Export Path]");
return;
}
try
{
var parameter = new Parameter();
parameter.CorpusPath = args[0];
parameter.Alpha = double.Parse(args[1]);
parameter.Beta = double.Parse(args[2]);
parameter.TopicCount = int.Parse(args[3]);
parameter.TotalIterationStep = int.Parse(args[4]);
parameter.ModelPath = args[5];
// parameter adjustment by topic count
//parameter.Alpha /= parameter.TopicCount;
parameter.LoadCorpus();
var lda = new LDA(parameter);
lda.Inference();
}
catch (Exception ex)
{
LogHelper.Log(ex);
}
Console.ReadKey();
}
private void buttonLoadModel_Click(object sender, EventArgs e)
{
if (modelFolderDialog.ShowDialog() == DialogResult.OK)
{
var path = modelFolderDialog.SelectedPath;
Parameter = path.Import<Parameter>();
LDAModel = path.Import<LDAModel>();
ModelHelper.ImportVoca(path);
UpdateTopicGridView();
}
}
public void UpdateTopicModel(LDAModel model, Parameter parameter, int topWordCount = 5)
{
_ldaModel = model;
_parameter = parameter;
_topWordCount = topWordCount;
var topicCount = _parameter.TopicCount;
// extract top words from each topic
_topicTopWordDist = new List<List<Tuple<int, double>>>(topicCount);
foreach (var topicId in Enumerable.Range(0, _parameter.TopicCount))
{
_topicTopWordDist.Add(
model.Phi[topicId]
.Select((elem, idx) => new Tuple<int, double>(idx, elem))
.OrderByDescending(e => e.Item2)
.Take(_topWordCount)
.ToList());
}
// extract top word set
_topWordSet = new HashSet<int>(_topicTopWordDist
.SelectMany(e => e.Select(t => t.Item1))
.Distinct()
.OrderBy(e => e));
// resize
var startX = 160;
var startY = 160;
var len = 30;
var width = len * (topicCount - 1);
var height = len * (_topWordSet.Count - 1);
var totalWidth = startX + width + 50;
var totalHeight = startY + height + 50;
Width = totalWidth;
Height = totalHeight;
}
public LDA(Parameter parameters)
{
_parameter = parameters;
_ldaModel = new LDAModel(_parameter);
}
public List<List<int>> Z; // topic assignment for each word in documents: | Documents | * | Word Count for each document |
#endregion Fields
#region Constructors
public LDAModel(Parameter parameter)
{
Parameter = parameter;
#region initialize LDA model hyper paramters
var docCount = Parameter.DocumentList.Count;
var vocaCount = WordManager.VocabularyCount;
var topicCount = Parameter.TopicCount;
// Z
Z = new List<List<int>>(docCount);
foreach (var document in Parameter.DocumentList.Select((document, idx) => new { Index = idx, WordCount = document.Count }))
Z.Add(Enumerable.Repeat(0, document.WordCount).ToList());
// NW
NW = ModelHelper.InitializeMatrix<int>(vocaCount, topicCount);
// ND
ND = ModelHelper.InitializeMatrix<int>(docCount, topicCount);
// NWCount
NWCount = ModelHelper.InitializeList<int>(topicCount);
// Theta
Theta = ModelHelper.InitializeMatrix<double>(docCount, topicCount);
// Phi
Phi = ModelHelper.InitializeMatrix<double>(topicCount, vocaCount);
#endregion
}
