public override ObjectClassificationResult PredictProba(double[] sarr)
{
var result = new ObjectClassificationResult();
// modifying data
for (int i = 0; i < sarr.Length; i++)
{
sarr[i] -= _avgs[i];
sarr[i] /= _vars[i] > 0 ? Math.Sqrt(_vars[i]) : 1;
}
var source = _trainLoader.Rows;
var tlist = new ConcurrentBag <Tuple <double, double> >();
List <Tuple <double, double> > taskList;
if (IsParallel)
{
taskList = (from row in source.AsParallel()
select ProceedRow(sarr, row)
).ToList();
}
else
{
taskList = (from row in source
select ProceedRow(sarr, row)
).ToList();
}
taskList.ForEach(t => tlist.Add(t));
var slist = tlist.OrderBy(t => t.Item1).ToArray();
var pcounts = new SortedDictionary <double, double>();
var sb = new StringBuilder();
for (int i = NeighborsOffset, ilen = 0; i < KNeighbors + NeighborsOffset; i++, ilen++)
{
var targ = slist[i].Item2;
if (!pcounts.ContainsKey(targ))
{
pcounts.Add(targ, 0);
}
pcounts[targ]++;
if (ilen < InfoLength)
{
sb.Append(slist[i].Item1.ToString("F08") + ';');
}
}
result.ObjectInfo = sb.ToString();
result.Probs = new double[_nclasses];
for (int i = 0; i < _nclasses; i++)
{
result.Probs[i] = pcounts.ContainsKey(i) ? pcounts[i] / KNeighbors : 0;
}
return(result);
}
/// <summary>
/// Predict probability for object
/// </summary>
/// <param name="sarr">array of double params</param>
/// <returns></returns>
public override ObjectClassificationResult PredictProba(double[] sarr)
{
var result = new ObjectClassificationResult();
var y = PredictCounts(sarr);
int cnt = _classifiers.Keys.Sum(id => _classifiers[id].CountTreesInBatch);
for (int i = 0; i < y.Length; i++)
{
y[i] /= cnt;
}
result.Probs = y;
return(result);
}
public override ObjectClassificationResult PredictProba(double[] sarr)
{
var result = new ObjectClassificationResult();
Node[] x = new Node[sarr.Length];
for (int j = 0; j < sarr.Length; j++)
{
x[j] = new Node();
x[j].Index = j + 1;
x[j].Value = sarr[j];
}
var ret = Prediction.PredictProbability(_model, x);
result.Probs = ret;
return(result);
}
/// <summary>
/// Predict probability for object
/// </summary>
/// <param name="sarr">array of double params</param>
/// <returns></returns>
public override ObjectClassificationResult PredictProba(double[] sarr)
{
var result = new ObjectClassificationResult();
var y = PredictCounts(sarr);
int cnt = _classifiers.Keys.Sum(id => _classifiers[id].CountTreesInBatch);
for (int i = 0; i < y.Length; i++)
y[i] /= cnt;
result.Probs = y;
return result;
}
public override ObjectClassificationResult PredictProba(double[] sarr)
{
var result = new ObjectClassificationResult();
// modifying data
for (int i = 0; i < sarr.Length; i++)
{
sarr[i] -= _avgs[i];
sarr[i] /= _vars[i] > 0 ? Math.Sqrt(_vars[i]) : 1;
}
var source = _trainLoader.Rows;
var tlist = new ConcurrentBag<Tuple<double, double>>();
List<Tuple<double, double>> taskList;
if (IsParallel)
taskList = (from row in source.AsParallel()
select ProceedRow(sarr, row)
).ToList();
else
taskList = (from row in source
select ProceedRow(sarr, row)
).ToList();
taskList.ForEach(t => tlist.Add(t));
var slist = tlist.OrderBy(t => t.Item1).ToArray();
var pcounts = new SortedDictionary<double, double>();
var sb = new StringBuilder();
for (int i= NeighborsOffset,ilen=0; i<KNeighbors + NeighborsOffset; i++,ilen++)
{
var targ = slist[i].Item2;
if (!pcounts.ContainsKey(targ))
pcounts.Add(targ, 0);
pcounts[targ]++;
if (ilen<InfoLength)
sb.Append(slist[i].Item1.ToString("F08")+';');
}
result.ObjectInfo = sb.ToString();
result.Probs = new double[_nclasses];
for (int i=0;i<_nclasses;i++)
{
result.Probs[i] = pcounts.ContainsKey(i) ? pcounts[i] / KNeighbors : 0;
}
return result;
}
public override ObjectClassificationResult PredictProba(double[] sarr)
{
var result = new ObjectClassificationResult();
Node[] x = new Node[sarr.Length];
for (int j = 0; j < sarr.Length; j++)
{
x[j] = new Node();
x[j].Index = j + 1;
x[j].Value = sarr[j];
}
var ret = Prediction.PredictProbability(_model, x);
result.Probs = ret;
return result;
}
