mirror of
https://github.com/hanzoai/jube-fork.git
synced 2026-10-30 18:33:33 +00:00
richard-churchman
cc0a2bee21
Included open-source Accord.net library code natively to the solution given the deprecated state of that project. Accord.net will be supported inside Jube for own needs (such as upgrading to .Net 7+). Accord.net libraries modified to compile to .Net 6. Modified Accord.net libraries to better support Newtonsoft Json.net serialisation. Included annotations in the Accord.net libraries to house important dependency statistics. Serialised json can be used independently of Jube to recall a Neural Network. Removal of Binary Serializer totally as a precursor to the projects support of .Net 7+. Extensive bug fixing in crucial functions of Neural Network training and recall. While this part of the system remains in Alpha, it is substantially more fit for purpose.
291 lines
8.8 KiB
C#
291 lines
8.8 KiB
C#
// Accord Machine Learning Library
|
|
// The Accord.NET Framework
|
|
// http://accord-framework.net
|
|
//
|
|
// Copyright © César Souza, 2009-2017
|
|
// cesarsouza at gmail.com
|
|
//
|
|
// This library is free software; you can redistribute it and/or
|
|
// modify it under the terms of the GNU Lesser General Public
|
|
// License as published by the Free Software Foundation; either
|
|
// version 2.1 of the License, or (at your option) any later version.
|
|
//
|
|
// This library is distributed in the hope that it will be useful,
|
|
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
// Lesser General Public License for more details.
|
|
//
|
|
// You should have received a copy of the GNU Lesser General Public
|
|
// License along with this library; if not, write to the Free Software
|
|
// Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
|
|
//
|
|
|
|
namespace Accord.MachineLearning.DecisionTrees.Rules
|
|
{
|
|
using Accord.Math;
|
|
using Accord.Statistics.Analysis;
|
|
using Accord.Statistics.Testing;
|
|
using System.Collections.Generic;
|
|
using System.Linq;
|
|
|
|
/// <summary>
|
|
/// Decision rule simplification algorithm.
|
|
/// </summary>
|
|
///
|
|
public class Simplification
|
|
{
|
|
DecisionSet decisionList;
|
|
double alpha = 0.05;
|
|
|
|
/// <summary>
|
|
/// Gets or sets the underlying hypothesis test
|
|
/// size parameter used to reject hypothesis.
|
|
/// </summary>
|
|
///
|
|
public double Alpha
|
|
{
|
|
get { return alpha; }
|
|
set { alpha = value; }
|
|
}
|
|
|
|
/// <summary>
|
|
/// Initializes a new instance of the <see cref="Simplification"/> class.
|
|
/// </summary>
|
|
///
|
|
/// <param name="list">The decision set to be simplified.</param>
|
|
///
|
|
public Simplification(DecisionSet list)
|
|
{
|
|
this.decisionList = list;
|
|
}
|
|
|
|
|
|
|
|
/// <summary>
|
|
/// Computes the reduction algorithm.
|
|
/// </summary>
|
|
///
|
|
/// <param name="inputs">A set of training inputs.</param>
|
|
/// <param name="outputs">The outputs corresponding to each of the inputs.</param>
|
|
///
|
|
/// <returns>The average error after the reduction.</returns>
|
|
///
|
|
public double Compute(double[][] inputs, int[] outputs)
|
|
{
|
|
int samples = outputs.Length;
|
|
|
|
bool[] actual = new bool[samples];
|
|
bool[] expected = new bool[samples];
|
|
|
|
DecisionRule[] list = decisionList.ToArray();
|
|
|
|
var antecedents = new HashSet<Antecedent>();
|
|
|
|
|
|
foreach (DecisionRule rule in list)
|
|
foreach (Antecedent antecedent in rule)
|
|
antecedents.Add(antecedent);
|
|
|
|
|
|
|
|
// 1. Eliminate unnecessary antecedents
|
|
for (int y = 0; y < decisionList.NumberOfClasses; y++)
|
|
{
|
|
for (int i = 0; i < outputs.Length; i++)
|
|
expected[i] = outputs[i] == y;
|
|
|
|
var unnecessary = new HashSet<Antecedent>();
|
|
|
|
/*foreach (var rule in list)
|
|
{
|
|
if (rule.Output != y)
|
|
continue;
|
|
*/
|
|
foreach (var antecedent in antecedents)
|
|
{
|
|
for (int i = 0; i < inputs.Length; i++)
|
|
actual[i] = antecedent.Match(inputs[i]);
|
|
|
|
if (CanEliminate(actual, expected))
|
|
unnecessary.Add(antecedent);
|
|
}
|
|
//}
|
|
|
|
foreach (var antecedent in unnecessary)
|
|
{
|
|
foreach (var rule in list)
|
|
{
|
|
if (rule.Output == y)
|
|
rule.Antecedents.Remove(antecedent);
|
|
}
|
|
}
|
|
}
|
|
|
|
bool[][] matches = new bool[list.Length][];
|
|
int[] counts = new int[list.Length];
|
|
for (int i = 0; i < matches.Length; i++)
|
|
{
|
|
DecisionRule rule = list[i];
|
|
|
|
matches[i] = new bool[outputs.Length];
|
|
for (int j = 0; j < inputs.Length; j++)
|
|
{
|
|
matches[i][j] = rule.Match(inputs[j]);
|
|
|
|
if (matches[i][j])
|
|
counts[i]++;
|
|
}
|
|
}
|
|
|
|
// double start = computeError(inputs, outputs, list);
|
|
|
|
for (int i = 0; i < list.Length; i++)
|
|
{
|
|
if (list[i] == null)
|
|
continue;
|
|
|
|
for (int j = 0; j < list.Length; j++)
|
|
{
|
|
if (list[i] == null)
|
|
break;
|
|
if (list[j] == null)
|
|
continue;
|
|
|
|
if (list[i].IsInconsistentWith(list[j]))
|
|
{
|
|
if (counts[i] > counts[j])
|
|
{
|
|
list[j] = null;
|
|
counts[j] = 0;
|
|
}
|
|
else
|
|
{
|
|
list[i] = null;
|
|
counts[i] = 0;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
list = list.Distinct(allowNulls: false);
|
|
|
|
List<DecisionRule> newList = new List<DecisionRule>(list);
|
|
|
|
// double middle = computeError(inputs, outputs, list);
|
|
|
|
// 2. Eliminate redundant rules from the set
|
|
|
|
for (int y = 0; y < decisionList.NumberOfClasses; y++)
|
|
{
|
|
for (int i = 0; i < outputs.Length; i++)
|
|
expected[i] = outputs[i] == y;
|
|
|
|
var unnecessary = new HashSet<DecisionRule>();
|
|
|
|
foreach (var rule in newList)
|
|
{
|
|
if (rule.Output != y)
|
|
continue;
|
|
|
|
for (int i = 0; i < inputs.Length; i++)
|
|
actual[i] = rule.Match(inputs[i]);
|
|
|
|
if (CanEliminate(actual, expected))
|
|
unnecessary.Add(rule);
|
|
}
|
|
|
|
foreach (var rule in unnecessary)
|
|
{
|
|
newList.Remove(rule);
|
|
}
|
|
}
|
|
|
|
|
|
// double final = computeError(inputs, outputs, newList);
|
|
|
|
decisionList.Clear();
|
|
decisionList.AddRange(newList);
|
|
|
|
// Compute new decision error
|
|
double newError = ComputeError(inputs, outputs);
|
|
|
|
return newError;
|
|
}
|
|
|
|
/// <summary>
|
|
/// Computes the average decision error.
|
|
/// </summary>
|
|
///
|
|
/// <param name="inputs">A set of input vectors.</param>
|
|
/// <param name="outputs">A set of corresponding output vectors.</param>
|
|
///
|
|
/// <returns>The average misclassification rate.</returns>
|
|
///
|
|
public double ComputeError(double[][] inputs, int[] outputs)
|
|
{
|
|
return computeError(inputs, outputs, decisionList);
|
|
}
|
|
|
|
private static double computeError(double[][] inputs, int[] outputs, IEnumerable<DecisionRule> rules)
|
|
{
|
|
int errors = 0;
|
|
for (int i = 0; i < outputs.Length; i++)
|
|
{
|
|
double expected = outputs[i];
|
|
double? actual = match(rules, inputs[i]);
|
|
|
|
if (expected != actual)
|
|
errors++;
|
|
}
|
|
|
|
return errors / (double)inputs.Length;
|
|
}
|
|
|
|
/// <summary>
|
|
/// Checks if two variables can be eliminated.
|
|
/// </summary>
|
|
///
|
|
public bool CanEliminate(bool[] actual, bool[] expected)
|
|
{
|
|
return CanEliminate(actual, expected, alpha);
|
|
}
|
|
|
|
/// <summary>
|
|
/// Checks if two variables can be eliminated.
|
|
/// </summary>
|
|
///
|
|
public static bool CanEliminate(bool[] actual, bool[] expected, double alpha)
|
|
{
|
|
var matrix = new ConfusionMatrix(actual, expected);
|
|
|
|
double maxExpectedFrequency = matrix.ExpectedValues.Max();
|
|
|
|
IHypothesisTest test;
|
|
|
|
if (maxExpectedFrequency > 10)
|
|
test = new ChiSquareTest(matrix, yatesCorrection: false) { Size = alpha };
|
|
|
|
else if (maxExpectedFrequency >= 5)
|
|
test = new ChiSquareTest(matrix, yatesCorrection: true) { Size = alpha };
|
|
|
|
else
|
|
test = new FisherExactTest(matrix) { Size = alpha };
|
|
|
|
return !test.Significant;
|
|
}
|
|
|
|
|
|
private static double? match(IEnumerable<DecisionRule> rules, double[] input)
|
|
{
|
|
foreach (var rule in rules)
|
|
{
|
|
if (rule.Match(input))
|
|
return rule.Output;
|
|
}
|
|
|
|
return null;
|
|
}
|
|
|
|
}
|
|
}
|