SVMKernel: Fix setParameter

lib/src/ExponentialRBF.cxx

@@ -63,7 +63,7 @@ Scalar ExponentialRBF::getSigma() const
   return sigma_;
 }
 
-void ExponentialRBF::setSigma(Scalar sigma)
+void ExponentialRBF::setSigma(const Scalar sigma)
 {
   sigma_ = sigma;
 }
@@ -77,6 +77,8 @@ Point ExponentialRBF::getParameter() const
 
 void ExponentialRBF::setParameter(const Point & parameter)
 {
+  if (parameter.getDimension() != 1)
+    throw InvalidArgumentException(HERE) << "ExponentialRBF expected a parameter of dimension 1";
   sigma_ = parameter[0];
 }
 

lib/src/NormalRBF.cxx

@@ -63,7 +63,7 @@ Scalar NormalRBF::getSigma() const
   return sigma_;
 }
 
-void NormalRBF::setSigma(Scalar sigma)
+void NormalRBF::setSigma(const Scalar sigma)
 {
   sigma_ = sigma;
 }
@@ -77,6 +77,8 @@ Point NormalRBF::getParameter() const
 
 void NormalRBF::setParameter(const Point & parameter)
 {
+  if (parameter.getDimension() != 1)
+    throw InvalidArgumentException(HERE) << "NormalRBF expected a parameter of dimension 1";
   sigma_ = parameter[0];
 }
 
@@ -91,8 +93,8 @@ Description NormalRBF::getParameterDescription() const
 /* Operator () */
 Scalar NormalRBF::operator() (const Point & x1, const Point & x2) const
 {
-  Point difference(x1 - x2);
-  Scalar value = exp(- difference.normSquare() / (2.0 * sigma_ * sigma_));
+  const Point difference(x1 - x2);
+  const Scalar value = exp(- difference.normSquare() / (2.0 * sigma_ * sigma_));
   return value;
 }
 

lib/src/PolynomialKernel.cxx

@@ -101,13 +101,16 @@ void PolynomialKernel::setConstant(Scalar constant)
 /* Accessor to the parameter used for cross-validation */
 Point PolynomialKernel::getParameter() const
 {
-  return {degree_};
+  return {degree_, linear_, constant_};
 }
 
 void PolynomialKernel::setParameter(const Point & parameter)
 {
+  if (parameter.getDimension() != 3)
+    throw InvalidArgumentException(HERE) << "PolynomialKernel expected a parameter of dimension 3";
   degree_ = parameter[0];
-  constant_ = parameter[1];
+  linear_ = parameter[1];
+  constant_ = parameter[2];
 }
 
 

lib/src/RationalKernel.cxx

@@ -77,6 +77,8 @@ Point RationalKernel::getParameter() const
 
 void RationalKernel::setParameter(const Point & parameter)
 {
+  if (parameter.getDimension() != 1)
+    throw InvalidArgumentException(HERE) << "RationalKernel expected a parameter of dimension 1";
   constant_ = parameter[0];
 }
 

lib/src/SigmoidKernel.cxx

@@ -55,8 +55,8 @@ SigmoidKernel * SigmoidKernel::clone() const
 String SigmoidKernel::__repr__() const
 {
   return OSS()  << "class=" << getClassName()
-         << "constant=" << constant_
-         << "linear=" << linear_;
+         << " constant=" << constant_
+         << " linear=" << linear_;
 }
 
 
@@ -66,7 +66,7 @@ Scalar SigmoidKernel::getLinear() const
   return linear_;
 }
 
-void SigmoidKernel::setLinear( Scalar linear )
+void SigmoidKernel::setLinear(const Scalar linear)
 {
   linear_ = linear;
 }
@@ -78,13 +78,15 @@ Scalar SigmoidKernel::getConstant() const
   return constant_;
 }
 
-void SigmoidKernel::setConstant( Scalar constant )
+void SigmoidKernel::setConstant(const Scalar constant)
 {
   constant_ = constant;
 }
 
 void SigmoidKernel::setParameter(const Point & parameter)
 {
+  if (parameter.getDimension() != 2)
+    throw InvalidArgumentException(HERE) << "SigmoidKernel expected a parameter of dimension 2";
   linear_ = parameter[0];
   constant_ = parameter[1];
 }
@@ -101,10 +103,10 @@ Description SigmoidKernel::getParameterDescription() const
 }
 
 /* Operator () */
-Scalar SigmoidKernel::operator() ( const Point & x1, const Point & x2 ) const
+Scalar SigmoidKernel::operator() (const Point & x1, const Point & x2) const
 {
-  Scalar dotProduct = x1.dot(x2);
-  Scalar value = tanh( linear_ * dotProduct + constant_ );
+  const Scalar dotProduct = x1.dot(x2);
+  const Scalar value = tanh( linear_ * dotProduct + constant_ );
   return value;
 }
 

lib/src/otsvm/ExponentialRBF.hxx

@@ -27,22 +27,16 @@ namespace OTSVM
 {
 
 
-
 /**
  * @class ExponentialRBF
  *
  * Implementation of the Exponential RBF kernel
  */
-
-
-
 class OTSVM_API ExponentialRBF
   : public SVMKernelImplementation
 {
   CLASSNAME
 public:
-
-
   /** Constructor with parameters */
   explicit ExponentialRBF(const OT::Scalar sigma = 1.0);
 
@@ -54,7 +48,7 @@ public:
 
   /** Sigma parameter accessor */
   virtual OT::Scalar getSigma() const;
-  virtual void setSigma(OT::Scalar sigma);
+  virtual void setSigma(const OT::Scalar sigma);
 
   /** Accessor to the parameter used for cross-validation */
   OT::Point getParameter() const override;

lib/src/otsvm/NormalRBF.hxx

@@ -49,7 +49,7 @@ public:
 
   /** Sigma parameter accessor */
   virtual OT::Scalar getSigma() const;
-  virtual void setSigma(OT::Scalar sigma);
+  virtual void setSigma(const OT::Scalar sigma);
 
   /** Accessor to the parameter used for cross-validation */
   OT::Point getParameter() const override;

lib/src/otsvm/SigmoidKernel.hxx

@@ -48,11 +48,11 @@ public:
 
   /* Linear term accessor */
   virtual OT::Scalar getLinear() const;
-  virtual void setLinear(OT::Scalar linear);
+  virtual void setLinear(const OT::Scalar linear);
 
   /* Constant term accessor */
   virtual OT::Scalar getConstant() const;
-  virtual void setConstant(OT::Scalar constant);
+  virtual void setConstant(const OT::Scalar constant);
 
   /* Parameters value and description accessor */
   OT::Point getParameter() const override;

python/test/t_SVMKernel_std.py

@@ -1,7 +1,9 @@
 #! /usr/bin/env python
 
+import openturns as ot
 import openturns.testing as ott
 import otsvm
+import os
 
 kernels = [
     otsvm.NormalRBF(2.0),
@@ -21,6 +23,7 @@
 V.append([x, y])
 
 for kernel in kernels:
+    print(kernel)
     for v in V:
         x, y = v
         print("x,y=", x, y)
@@ -43,3 +46,22 @@
 
         kxydx2 = kernel.partialHessian(x, y)
         print("d2kernel/(dx_i*dx_j)(x,y)=", repr(kxydx2))
+
+    # parameter accessor
+    param = kernel.getParameter()
+    kernel.setParameter(param)
+
+    # serialization
+    if ot.PlatformInfo.HasFeature("libxml2"):
+        study = ot.Study()
+        fname = "study_kernel.xml"
+        study.setStorageManager(ot.XMLStorageManager(fname))
+        study.add("kernel", kernel)
+        study.save()
+        study = ot.Study()
+        study.setStorageManager(ot.XMLStorageManager(fname))
+        study.load()
+        kernel = otsvm.SVMKernel()
+        study.fillObject("kernel", kernel)
+        print(kernel)
+        os.remove(fname)