public bool Initialize(Header header)
{
// Ensure we have reproducible results
Generator.Seed = 0;
learner = new B.NaiveBayesLearning()
{
};
Initialized = true;
return(Initialized);
}
public void run_shit()
{
naiveBayes = new Accord.MachineLearning.Bayes.NaiveBayesLearning();
load_dictionary();
//load_files();
//load_ball_training();
//load_empty_training();
//load_error_training();
//Arffgen arffgen = new Arffgen(trainingInput,trainingOutput,arrayCounter);
load_model();
build_model();
//save_model();
}
public double Estimate(double[][] inputs, int[] outputs, bool empirical = true)
{
var teacher = new NaiveBayesLearning <TDistribution>()
{
Model = this
};
#if DEBUG
teacher.ParallelOptions.MaxDegreeOfParallelism = 1;
#endif
teacher.Empirical = empirical;
NaiveBayes <TDistribution, double> result = teacher.Learn(inputs, outputs);
base.Distributions = result.Distributions;
this.Priors = result.Priors;
return(new ZeroOneLoss(outputs)
{
Mean = true
}.Loss(Decide(inputs)));
}
public double Estimate <TOptions>(double[][] inputs, int[] outputs,
bool empirical = true, TOptions options = null)
where TOptions : class, IFittingOptions, new()
{
var teacher = new NaiveBayesLearning <TDistribution, TOptions>()
{
Model = this
};
#if DEBUG
teacher.ParallelOptions.MaxDegreeOfParallelism = 1;
#endif
teacher.Empirical = empirical;
teacher.Options.InnerOption = options;
NaiveBayes <TDistribution, double> result = teacher.Learn(inputs, outputs);
base.Distributions = result.Distributions;
this.Priors = result.Priors;
return(new ZeroOneLoss(outputs)
{
Mean = true
}.Loss(Decide(inputs)));
}
public double Estimate(int[][] inputs, int[] outputs,
bool empirical = true, double regularization = 1e-5)
{
var teacher = new NaiveBayesLearning()
{
Model = this
};
teacher.Empirical = empirical;
teacher.Options.InnerOption.Regularization = regularization;
teacher.Options.InnerOption.Minimum = 0;
#if DEBUG
teacher.ParallelOptions.MaxDegreeOfParallelism = 1;
#endif
var result = teacher.Learn(inputs, outputs);
var b = result as NaiveBayes <GeneralDiscreteDistribution, int>;
base.Distributions = b.Distributions;
this.Priors = b.Priors;
this.symbols = result.symbols;
return(new ZeroOneLoss(outputs)
{
Mean = true
}.Loss(Decide(inputs)));
}