/// <summary> /// Gets the local minimums distribution. /// </summary> /// <param name="dmFieldData">The dm field data.</param> /// <param name="dimensionNumber">The dimension number: /// 1 - rows (angle) /// 2 - columns (distance) /// </param> /// <returns>DenseMatrix.</returns> //public static DenseMatrix GetLocalMinimumsDistribution(DenseMatrix dmFieldData, PointD sunCenterPoint, PointD imageCenterPoint, double imageRadius, int imageHeight, double imageCircleCropFactor = 0.9d, int dimensionNumber = 1) public static List <Point3D> GetLocalMinimumsDistribution(DenseMatrix dmFieldData, RoundData sunDiskData, RoundData imageRoundData, int imageHeight, double imageCircleCropFactor = 0.9d) { // DenseMatrix dmFieldminimumsData = DenseMatrix.Create(dmFieldData.RowCount, dmFieldData.ColumnCount, 0.0d); List <Point3D> lRetPoints = new List <Point3D>(); double imageRadius = imageRoundData.DRadius; PointD imageCenterPoint = imageRoundData.pointDCircleCenter(); PointPolar imageCenterPointRelatedToSunCenter = new PointPolar(imageCenterPoint - sunDiskData.pointDCircleCenter(), true); double distanceSunCenterToImageCenter = PointD.Distance(imageCenterPoint, sunDiskData.pointDCircleCenter()); #region // obsolete //if (dimensionNumber == 1) //{ #endregion // obsolete for (int i = 0; i < dmFieldData.RowCount; i++) { bool itsTheCropCase = false; //если направлени на кроп кадра - то не берем в расмотрение double currentAngle = ((double)i / (double)(dmFieldData.RowCount - 1)) * 2.0d * Math.PI; LineDescription2D line, lineMargin; if (currentAngle < Math.PI) { //верхняя половина, смотрим направление на y=0.0d line = new LineDescription2D(sunDiskData.pointDCircleCenter(), new Vector2D(Math.Cos(currentAngle), -Math.Sin(currentAngle))); lineMargin = new LineDescription2D(new PointD(0.0d, 0.0d), new Vector2D(1.0d, 0.0d)); } else { line = new LineDescription2D(sunDiskData.pointDCircleCenter(), new Vector2D(Math.Cos(currentAngle), Math.Sin(currentAngle))); lineMargin = new LineDescription2D(new PointD(0.0d, imageHeight), new Vector2D(1.0d, 0.0d)); } PointD crossPointD = LineDescription2D.CrossPoint(line, lineMargin); if (crossPointD.Distance(imageCenterPoint) < imageRadius) { itsTheCropCase = true; } #region // obsolete //double yMargin = 0.0d; //double xMargin = sunCenterPoint.X + (yMargin - sunCenterPoint.Y) / Math.Tan(currentAngle); //double dx = xMargin - imageCenterPoint.X; //double dy = yMargin - imageCenterPoint.Y; //if (Math.Sqrt(dx * dx + dy * dy) < imageRadius) itsTheCropCase = true; #endregion // obsolete #region //obsolete //else //{ // //нижняя половина, смотрим направление на y=imageHeight // double yMargin = (double)imageHeight; // double xMargin = sunCenterPoint.X + (yMargin - sunCenterPoint.Y) / Math.Tan(currentAngle); // double dx = xMargin - imageCenterPoint.X; // double dy = yMargin - imageCenterPoint.Y; // if (Math.Sqrt(dx * dx + dy * dy) < imageRadius) itsTheCropCase = true; //} #endregion //obsolete //Если слишком близко к краю изображения - тоже исключаем. Минимум должен лежать не ближе, например, 1/15 //DenseMatrix dmSlicedDataMatrix = (DenseMatrix)dmFieldData.SubMatrix(i, 1, 0, dmFieldData.ColumnCount); DenseVector dvRowDataVector = (DenseVector)dmFieldData.EnumerateRows().ElementAt(i); #region // debug plotting //dvRowDataVector.SaveVectorDataAsImagePlot( // "D:\\_gulevlab\\SkyImagesAnalysis_appData\\patent-samples\\result.2015-03-24\\img-2014-09-20T16-03-58devID1\\dvRowDataVector-plot-image-" + // i.ToString("D03") + "-step1.png"); #endregion // debug plotting dvRowDataVector.MapIndexedInplace((idx, x) => ((x == 0.0d) || (idx < sunDiskData.DRadius * 1.5d)) ? (1.0d) : (x)); #region // debug plotting //dvRowDataVector.SaveVectorDataAsImagePlot( // "D:\\_gulevlab\\SkyImagesAnalysis_appData\\patent-samples\\result.2015-03-24\\img-2014-09-20T16-03-58devID1\\dvRowDataVector-plot-image-" + // i.ToString("D03") + "-step2.png"); #endregion // debug plotting double phiFromImageCenterToDirection = imageCenterPointRelatedToSunCenter.Phi - currentAngle; double distanceToImageMargin = distanceSunCenterToImageCenter * Math.Cos(phiFromImageCenterToDirection) + Math.Sqrt(imageRadius * imageRadius - distanceSunCenterToImageCenter * distanceSunCenterToImageCenter * Math.Sin(phiFromImageCenterToDirection) * Math.Sin(phiFromImageCenterToDirection)); dvRowDataVector.MapIndexedInplace( (idx, x) => ((double)idx / distanceToImageMargin >= imageCircleCropFactor) ? (1.0d) : (x)); #region // debug plotting //dvRowDataVector.SaveVectorDataAsImagePlot( // "D:\\_gulevlab\\SkyImagesAnalysis_appData\\patent-samples\\result.2015-03-24\\img-2014-09-20T16-03-58devID1\\dvRowDataVector-plot-image-" + // i.ToString("D03") + "-step3.png"); #endregion // debug plotting double minValue = dvRowDataVector.Minimum(); int minValueIndex = dvRowDataVector.MinimumIndex(); //if (!itsTheCropCase) dmFieldminimumsData[i, minValueIndex] = minValue; if ((!itsTheCropCase) && ((double)minValueIndex > sunDiskData.DRadius)) { lRetPoints.Add(new Point3D(currentAngle, minValueIndex, minValue)); } else { continue; } } #region // obsolete //} //else if (dimensionNumber == 2) //{ // for (int i = 0; i < dmFieldData.ColumnCount; i++) // { // DenseMatrix dmSlicedDataMatrix = (DenseMatrix)dmFieldData.SubMatrix(0, dmFieldData.RowCount, i, 1); // DenseVector dvSlicedDataVector = DenseVector.OfEnumerable(dmSlicedDataMatrix.Values); // dvSlicedDataVector.MapInplace(new Func<double, double>(x => (x == 0.0d) ? (1.0d) : (x))); // double minValue = dvSlicedDataVector.Minimum(); // int minValueIndex = dvSlicedDataVector.MinimumIndex(); // dmFieldminimumsData[minValueIndex, i] = minValue; // } //} #endregion // obsolete //return dmFieldminimumsData; return(lRetPoints); }
public void Train(DenseMatrix X, DenseVector d, DenseVector Kd) { int R = X.RowCount; int N = X.ColumnCount; int U = 0; //the number of neurons in the structure var c = new DenseMatrix(R, 1); var sigma = new DenseMatrix(R, 1); var Q = new DenseMatrix((R + 1), (R + 1)); var O = new DenseMatrix(1, (R + 1)); var pT_n = new DenseMatrix((R + 1), 1); double maxPhi = 0; int maxIndex; var Psi = new DenseMatrix(N, 1); Console.WriteLine("Running..."); //for each observation n in X for (int i = 0; i < N; i++) { Console.WriteLine(100 * (i / (double)N) + "%"); var x = new DenseVector(R); X.Column(i, x); //if there are neurons in structure, //update structure recursively. if (U == 0) { c = (DenseMatrix)x.ToColumnMatrix(); sigma = new DenseMatrix(R, 1, SigmaZero); U = 1; Psi = CalculatePsi(X, c, sigma); UpdateStructure(X, Psi, d, ref Q, ref O); pT_n = (DenseMatrix) (CalculateGreatPsi((DenseMatrix)x.ToColumnMatrix(), (DenseMatrix)Psi.Row(i).ToRowMatrix())) .Transpose(); } else { StructureRecurse(X, Psi, d, i, ref Q, ref O, ref pT_n); } bool KeepSpinning = true; while (KeepSpinning) { //Calculate the error and if-part criteria double ee = pT_n.Multiply(O)[0, 0]; double approximationError = Math.Abs(d[i] - ee); DenseVector Phi; double SumPhi; CalculatePhi(x, c, sigma, out Phi, out SumPhi); maxPhi = Phi.Maximum(); maxIndex = Phi.MaximumIndex(); if (approximationError > delta) { if (maxPhi < threshold) { var tempSigma = new DenseVector(R); sigma.Column(maxIndex, tempSigma); double minSigma = tempSigma.Minimum(); int minIndex = tempSigma.MinimumIndex(); sigma[minIndex, maxIndex] = k_sigma * minSigma; Psi = CalculatePsi(X, c, sigma); UpdateStructure(X, Psi, d, ref Q, ref O); var psi = new DenseVector(Psi.ColumnCount); Psi.Row(i, psi); pT_n = (DenseMatrix) CalculateGreatPsi((DenseMatrix)x.ToColumnMatrix(), (DenseMatrix)psi.ToRowMatrix()) .Transpose(); } else { //add a new neuron and update strucutre double distance = 0; var cTemp = new DenseVector(R); var sigmaTemp = new DenseVector(R); //foreach input variable for (int j = 0; j < R; j++) { distance = Math.Abs(x[j] - c[j, 0]); int distanceIndex = 0; //foreach neuron past 1 for (int k = 1; k < U; k++) { if ((Math.Abs(x[j] - c[j, k])) < distance) { distanceIndex = k; distance = Math.Abs(x[j] - c[j, k]); } } if (distance < Kd[j]) { cTemp[j] = c[j, distanceIndex]; sigmaTemp[j] = sigma[j, distanceIndex]; } else { cTemp[j] = x[j]; sigmaTemp[j] = distance; } } //end foreach c = (DenseMatrix)c.InsertColumn(c.ColumnCount - 1, cTemp); sigma = (DenseMatrix)sigma.InsertColumn(sigma.ColumnCount - 1, sigmaTemp); Psi = CalculatePsi(X, c, sigma); UpdateStructure(X, Psi, d, ref Q, ref O); U++; KeepSpinning = false; } } else { if (maxPhi < threshold) { var tempSigma = new DenseVector(R); sigma.Column(maxIndex, tempSigma); double minSigma = tempSigma.Minimum(); int minIndex = tempSigma.MinimumIndex(); sigma[minIndex, maxIndex] = k_sigma * minSigma; Psi = CalculatePsi(X, c, sigma); UpdateStructure(X, Psi, d, ref Q, ref O); var psi = new DenseVector(Psi.ColumnCount); Psi.Row(i, psi); pT_n = (DenseMatrix) CalculateGreatPsi((DenseMatrix)x.ToColumnMatrix(), (DenseMatrix)psi.ToRowMatrix()) .Transpose(); } else { KeepSpinning = false; } } } } out_C = c; out_O = O; out_Sigma = sigma; Console.WriteLine("Done."); }
// Решение двойственным симплекс-методом public DenseVector Solve() { List <int> Jplus = null; // Jн+ List <int> Jminus = null; // Jн- while (true) { J.Sort(); var Ab = A.SelectColumns(J); var Abi = Ab.Inverse(); var Cb = C.Select(J); var E = DenseMatrix.Identity(M); Console.WriteLine(String.Join(" ", J)); var nn = new DenseVector(N); var final = true; var jk = -1; // Расчет Δ var delta = (DenseVector)(Cb * Abi); delta *= A; delta -= C; // Начальные значения Jн+ (по Δ) if (Jplus == null) { Jplus = new List <int> (); Jminus = new List <int> (); foreach (int j in Jn) { if (delta [j] >= 0) { Jplus.Add(j); } else { Jminus.Add(j); } } } // Расчет ℵ for (int i = 0; i < N; i++) { if (!J.Contains(i)) { if (Jplus.Contains(i)) { nn [i] = DL [i]; } else { nn [i] = DR [i]; } } } var sum = DenseVector.Create(M, (i) => 0); foreach (int i in Jn) { sum += (DenseVector)(A.Column(i) * nn [i]); } var nValues = (DenseMatrix)Abi * (B - sum); for (int i = 0; i < N; i++) { if (J.Contains(i)) { nn [i] = nValues [J.IndexOf(i)]; // Проверка ℵ по ограничениям var fits = nn [i] >= DL [i] && nn [i] <= DR [i]; final &= fits; if (!fits && jk == -1) { // Запоминаем jk jk = i; } } } // Если ℵ подходит, то завершаем if (final) { return(nn); } // Расчет μ var mu = new DenseVector(N); mu [jk] = (nn [jk] < DL [jk] ? 1 : -1); var dy = mu [jk] * E.Column(J.IndexOf(jk)) * Abi; for (int i = 0; i < N; i++) { if (!J.Contains(i) && i != jk) { mu [i] = dy * A.Column(i); } } // Расчет σ var sigma = new DenseVector(N); for (int i = 0; i < N; i++) { if (Jn.Contains(i)) { if ((Jplus.Contains(i) && mu [i] < 0) || (Jminus.Contains(i) && mu [i] > 0)) { sigma [i] = -delta [i] / mu [i]; } else { sigma [i] = double.PositiveInfinity; } } else { sigma [i] = double.PositiveInfinity; } } // Поиск σ0 var j0 = sigma.MinimumIndex(); var sigma0 = sigma [j0]; if (sigma0 == double.PositiveInfinity) { // Нет решений return(null); } // Обновляем базис и Jн+/Jн- J.Remove(jk); J.Add(j0); if (Jplus.Contains(j0)) { Jplus.Remove(j0); } if (mu [jk] == 1) { Jplus.Add(jk); } // Пересчет Jн- Jminus.Clear(); foreach (var j in Jn) { if (!Jplus.Contains(j)) { Jminus.Add(j); } } } }