public MetricSimilarity(MelodySequence seq, IMetric metric, SimilarityType type = SimilarityType.Cosine)
{
this.metrics = new IMetric[]{metric};
this.target = seq.ToArray();
this.type = type;
this.target_metrics = new Dictionary<Pair, float>[] { this.metrics[0].Generate(this.target) };
}
public MetricSimilarity(MelodySequence seq, IMetric[] metrics, SimilarityType type = SimilarityType.Cosine)
{
this.metrics = metrics;
this.target = seq.ToArray();
target_metrics = new Dictionary<Pair, float>[metrics.Length];
this.type = type;
for (int i = 0; i < metrics.Length; i++)
target_metrics[i] = this.metrics[i].Generate(this.target);
}
private void calculateButton_Click(object sender, EventArgs e)
{
string[] items = new string[fileListBox.Items.Count];
for (int i = 0; i < items.Length; i++)
{
items[i] = fileListBox.Items[i].ToString();
}
SimilarityType type = SimilarityType.Euclidean;
Enum.TryParse <SimilarityType>(similarityTypeComboBox.SelectedValue.ToString(), out type);
resultTextBox.Text = Ensemble.Similarity(type, items).ToString("0.00000");
}
public static MetricSimilarity GenerateMetricSimilarityMulti(IEnumerable<Composition> comps, IMetric[] metrics, SimilarityType type = SimilarityType.Cosine)
{
List<Note> notes = new List<Note>();
foreach(var comp in comps)
{
if (comp.Tracks.Count < 1)
continue;
var seq = comp.Tracks[0].GetMainSequence() as MelodySequence;
foreach (var n in seq.Notes)
notes.Add(n);
}
return new MetricSimilarity(new MelodySequence(notes), metrics, type);
}
public static string GetSimilarityTypeDescription(this SimilarityType methodType)
{
// get the field
var field = methodType.GetType().GetField(methodType.ToString());
var customAttributes = field.GetCustomAttributes(typeof(DescriptionAttribute), false);
if (customAttributes.Count() > 0)
{
return((customAttributes.ElementAt(0) as DescriptionAttribute).Description);
}
else
{
return(methodType.ToString());
}
}
public void ShowModelInfo()
{
Logger.WriteLine($"Encoder embedding dim: '{EncoderEmbeddingDim}'");
Logger.WriteLine($"Decoder embedding dim: '{DecoderEmbeddingDim}'");
Logger.WriteLine($"Encoder layer depth: '{EncoderLayerDepth}'");
Logger.WriteLine($"Decoder layer depth: '{DecoderLayerDepth}'");
Logger.WriteLine($"Encoder type: '{EncoderType}'");
Logger.WriteLine($"Decoder type: '{DecoderType}'");
Logger.WriteLine($"Hidden layer dim: '{HiddenDim}'");
Logger.WriteLine($"Enable segment embeddings: '{EnableSegmentEmbeddings}'");
Logger.WriteLine($"Enable shared embeddings: '{SharedEmbeddings}'");
Logger.WriteLine($"Enable tag embeddings: '{EnableTagEmbeddings}'");
Logger.WriteLine($"Multi-head size: '{MultiHeadNum}'");
Logger.WriteLine($"Pointer Generator: '{PointerGenerator}'");
if (!SimilarityType.IsNullOrEmpty())
{
Logger.WriteLine($"Similarity Type: '{SimilarityType}'");
}
if (SrcVocab != null)
{
Logger.WriteLine($"Source vocabulary size: '{SrcVocab.Count}'");
}
if (TgtVocab != null)
{
Logger.WriteLine($"Target vocabulary size: '{TgtVocab.Count}'");
}
if (ClsVocabs != null)
{
Logger.WriteLine($"The number of CLS vocabularies: '{ClsVocabs.Count}' ");
for (int i = 0; i < ClsVocabs.Count; i++)
{
Logger.WriteLine($"CLS vocabulary {i} size: {ClsVocabs[ i ].Count}");
}
}
}
public static double Similarity(SimilarityType type, params string[] files)
{
REngine.SetEnvironmentVariables();
// There are several options to initialize the engine, but by default the following suffice:
REngine engine = REngine.GetInstance();
engine.Evaluate("library(clue)");
string ensemble_statement = "ensemble = cl_ensemble(";
for (int i = 0; i < files.Length; i++)
{
string statement = string.Empty;
statement += $"array = read.csv(\"{files[i].Replace("\\", "\\\\")}\", header = FALSE)$V1\r\n";
statement += "if(is.na(array[1])){";
statement += $" P{i} = as.cl_partition(read.csv(\"{files[i].Replace("\\", "\\\\")}\", header = TRUE)$V1)";
statement += "}else{";
statement += $" P{i} = as.cl_partition(array)";
statement += "}";
engine.Evaluate(statement);
ensemble_statement += $"P{i},";
}
ensemble_statement = ensemble_statement.TrimEnd(new char[] { ',' }) + ")";
engine.Evaluate(ensemble_statement);
engine.Evaluate("ag=cl_agreement(ensemble)");
var ag = engine.Evaluate("ag");
Console.WriteLine(ag.ToString());
var mean = engine.Evaluate($"mean(cl_agreement(ensemble, method = \"{type}\"))").AsVector();
double result = (double)mean[0];
// you should always dispose of the REngine properly.
// After disposing of the engine, you cannot reinitialize nor reuse it
//engine.Dispose();
return(result);
}
private async Task <List <TrademarkDto> > ParseServerResponse(string response, SimilarityType type)
{
var tradeMarkCollection = new List <TrademarkDto>();
if (!string.IsNullOrEmpty(response))
{
var results = JsonConvert.DeserializeObject <dynamic>(response);
foreach (var result in results.results)
{
int.TryParse(result.barcode.ToString(), out int supervisorBarcode);
var protDoc = _executor.GetQuery <GetProtectionDocsByBarcodeQuery>()
.Process(q => q.Execute(supervisorBarcode));
if (protDoc != null)
{
var tdmDet = _mapper.Map <TrademarkDto>(protDoc);
switch (type)
{
case SimilarityType.Image:
tdmDet.ImageSimilarity = result.similarity;
break;
case SimilarityType.Phonetic:
tdmDet.PhonSimilarity = result.similarity;
break;
case SimilarityType.Semantic:
tdmDet.SemSimilarity = result.similarity;
break;
}
byte[] responseImage = protDoc.Image;
tdmDet.PreviewImage =
responseImage != null ? $"data:image/png;base64,{Convert.ToBase64String(responseImage)}" : null;
tradeMarkCollection.Add(tdmDet);
}
}
}
return(tradeMarkCollection);
}