Class NNGenFirm

java.lang.Object
  |
  +--Firm
        |
        +--NNGenFirm

public class NNGenFirm

extends Firm

NNGenFirm.java
Copyright (c) Murat Yildizoglu, 2000-2001
Permission to use, copy, modify, and distribute this
software and its documentation for NON-COMMERCIAL purposes
and without fee is hereby granted provided that this
copyright notice appears in all copies.
*************************************************************
An evolutionnary firm inspired by Nelson and Winter (1982) but enriched by genetic algorithm and artificial neural network based R&D investment

Version:

1 August 2001

Author:

Murat Yildizoglu

See Also:

Firm, genetic, ANN

Fields inherited from class Firm

type

Constructor Summary

NNGenFirm(double K0, double prod0, double cost0, double d0, double sigma_in0, double delta0, double rd0, int g0, int c0, double xover0, double mutate0, int nxover0, boolean elitism0, double r0, double rd_min0, int numInputs0, int numHiddens0, int numOutputs0, int numEpochs0, int numPeriods0)
Initializes and computes the firm's caracteristics given the caracteristics of the industry to which it belongs.

Method Summary

void	clean() Frees the Genetic algorithm and the ANN for garbage collection
int	getChrDecode(int cc) Decodes the actual chromosome in order to obtain the corresponding RD ratio.
double	getChrFitness(int cd) Provies the fitness of a particular chromosome
double	getDFittness() Returns the variation of the profit rate
double	getProfit() Total profit of the firm profit_rate*K
void	learn(int date) Train the Neural Network with new inputs and observed outputs.
void	setenv(java.lang.String S)
void	setrd(int t) Compute RD following a GA rule where the fittness is provided by the expected profit rate given by the ANN.
java.lang.String	showFirm() Formats the output for one GenFirm.

Methods inherited from class Firm

computeprofit, getdprim, getdprin, getinvestRD, getKapital, getprod, getProfitRate, getrd, imitate, innovate, investK, investRD, isactive, saveprofit, supply

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

NNGenFirm

public NNGenFirm(double K0,
                 double prod0,
                 double cost0,
                 double d0,
                 double sigma_in0,
                 double delta0,
                 double rd0,
                 int g0,
                 int c0,
                 double xover0,
                 double mutate0,
                 int nxover0,
                 boolean elitism0,
                 double r0,
                 double rd_min0,
                 int numInputs0,
                 int numHiddens0,
                 int numOutputs0,
                 int numEpochs0,
                 int numPeriods0)

Initializes and computes the firm's caracteristics given the caracteristics of the industry to which it belongs.

Parameters:

K0 - Initial capital of the firm

prod0 - Initial productivity of the firm

cost0 - Unit using cost of capital

d0 - Calibration parameter for the innovation process

sigma_in0 - Standard deviation of innovative draws

delta0 - Depreciation rate of capital

rd0 - NA

g0 - Number of genes

c0 - Number of chromosomes

xover0 - Crossover rate

mutate0 - Mutations rate

nxover0 - Number of crossover points

elitism0 - Elitist GA (true) or not (false)

r0 - Interest rate in the economy

rd_min0 - Minimal RD invetment rate

numInputs0 - number of the inputs nodes

numHiddens0 - number of the hidden nodes

numOutputs0 - number of the outputs nodes

numEpochs0 - Number of training periods (ANN) in each time period

numPeriods0 - Number of periods of the simulation

Method Detail

learn

public void learn(int date)

Train the Neural Network with new inputs and observed outputs. The variables that will bu used as inputs and outputs for the ANN must be fixed in this method.

Overrides:

learn in class Firm

Parameters:

date - The index of the last active element of observations used during training

See Also:

NNeco

setrd

public void setrd(int t)

Compute RD following a GA rule where the fittness is provided by the expected profit rate given by the ANN. To evaluate the fittness of each rule, a new inputs vector with this value is sent to the ANN to get the corresponding expectation. When all rules are evaluated, the GA is used to generate a new population of rules for the next period. The rule with the highest expectation is used in the actual period.

Overrides:

setrd in class Firm

Parameters:

t - : the time period

getProfit

public double getProfit()

Total profit of the firm profit_rate*K

Overrides:

getProfit in class Firm

getChrDecode

public int getChrDecode(int cc)

Decodes the actual chromosome in order to obtain the corresponding RD ratio.

Overrides:

getChrDecode in class Firm

Parameters:

cc - The chromosome to decode

getDFittness

public double getDFittness()

Returns the variation of the profit rate

Overrides:

getDFittness in class Firm

Following copied from class: Firm

See Also:

GenFirm

getChrFitness

public double getChrFitness(int cd)

Provies the fitness of a particular chromosome

Overrides:

getChrFitness in class Firm

Parameters:

cd - The chromosome for which the fitness is asked

clean

public void clean()

Frees the Genetic algorithm and the ANN for garbage collection

Overrides:

clean in class Firm

See Also:

genetic

showFirm

public java.lang.String showFirm()

Formats the output for one GenFirm. It is used in LearnInd.

Overrides:

showFirm in class Firm

setenv

public void setenv(java.lang.String S)

Overrides:

setenv in class Firm