Home  
  Download  
  Publications  
  Examples  
  Contributions  
  Links  
  Contact  

Download

The Current Release is Flood 2. It includes:
  • A perceptron neuron model with:
    • Logistic (sigmoid) activation function.
    • Hyperbolic tangent (sigmoid) activation function.
    • Linear activation function
  • A multilayer perceptron network architecture with:
    • An arbitrary number of hidden layers.
    • Pre and postprocessing methods:
      • Mean and standard deviation.
      • Minimum and maximum.
    • Independent parameters.
    • Boundary conditions.
    • Lower and upper bounds. 
  • Several objective functionals for variational problems:
    • Data modeling problems:
      • Sum squared error.
      • Mean squared error.
      • Root mean squared error
      • Minkowski error.
      • Normalized squared error.
      • Regularized Minkowski error.
    • Examples of classical problems in the calculus of variations:
      • Geodesic problem.
      • Brachistochrone problem.
      • Catenary problem.
      • Isoperimetric problem
    • Examples of optimal control problems:
      • Car problem.
      • Aircraft landing problem.
      • Fed batch fermenter problem
    • Examples of inverse problems:
      • Disolution modeling of aluminium alloys
    • Examples of optimal shape design:
      • Minimum drag problem.
    • Workaround examples for function optimization problems:
      • De Jong’s function.
      • Rosenbrock’s function.
      • Rastrigin’s function.
      • Schwefel’s function.
      • Easom’s function.
      • Six-hump camel back function.
      • Plane-cylinder.
  • Several training algorithms to optimize the objective function:
    • Random search.
    • Gradient descent.
      • Training rate methods:
        • Fixed.
        • Golden section.
        • Brent’s method.
    • Conjugate gradient:
      • Training direction methods:
        • Polak-Ribiere.
        • Fletcher-Reeves.
      • Training rate methods:
        • Fixed.
        • Golden section.
        • Brent’s method.
    • Newton method.
    • Quasi-Newton method:
      • Inverse Hessian approximation methods:
        • DFP.
        • BFGS.
      • Training rate methods:
        • Fixed.
        • Golden section.
        • Brent’s method.
    • Evolutionary algorithm:
      • Fitness assignment methods:
        • Linear ranking.
      • Selection methods:
        • Roulette wheel.
        • Stochastic universal sampling.
      • Recombination methods:
        • Line.
        • Intermediate.
      • Mutation methods:
        • Uniform.
        • Normal.
  • Different Utilities:
    • Vector class.
    • Matrix class.
    • Input-target data set.
      • Pre and postprocessing methods:
        • Mean and standard deviation.
        • Minimum and maximum.
    • Linear regression analysis.
    • Integration of functions:
      • Trapezoid rule.
      • Simpson’s method.
    • Ordinary differential equations:
      • Runge-Kutta method.
      • Runge-Kutta-Fehlberg method.  


The package comes with the following documentation:


Flood is written in ANSI C++ and should work on any compiler. So far it has been tested in Linux with gcc and in Windows with Microsoft Visual C++.