在 Accord.Net 中使用 AdaBoost(Boosting)

Question

我正在尝试在 Accord.Net 中使用 adaboost（或 boosting）。我尝试了https://github.com/accord-net/framework/wiki/Classification 给出的决策树示例的一个版本，它适用于以下代码：

'' Creates a matrix from the entire source data table
Dim data As DataTable = CType(DataView.DataSource, DataTable)

'' Create a new codification codebook to 
'' convert strings into integer symbols
Dim codebook As New Codification(data)

'' Translate our training data into integer symbols using our codebook:
Dim symbols As DataTable = codebook.Apply(data)
Dim inputs As Double()() = symbols.ToArray(Of Double)("Outlook", "Temperature", "Humidity", "Wind")
Dim outputs As Integer() = symbols.ToArray(Of Integer)("PlayTennis")

'' Gather information about decision variables
Dim attributes() As DecisionVariable = {New DecisionVariable("Outlook", 3), New DecisionVariable("Temperature", 3), _
    New DecisionVariable("Humidity", 2), New DecisionVariable("Wind", 2)}

Dim classCount As Integer = 2 '' 2 possible output values for playing tennis: yes or no

''Create the decision tree using the attributes and classes
tree = New DecisionTree(attributes, classCount)

'' Create a new instance of the ID3 algorithm
Dim Learning As New C45Learning(tree)

'' Learn the training instances!
Learning.Run(inputs, outputs)

Dim aa As Integer() = codebook.Translate("D1", "Rain", "Mild", "High", "Weak")

Dim ans As Integer = tree.Compute(aa)

Dim answer As String = codebook.Translate("PlayTennis", ans)

现在我想添加此代码以在更复杂的示例中使用 adaboost 或 boosting。我通过在上面的代码中添加以下内容来尝试以下操作：

Dim Booster As New Boost(Of DecisionStump)()

Dim Learn As New AdaBoost(Of DecisionStump)(Booster)
Dim weights(inputs.Length - 1) As Double
For i As Integer = 0 To weights.Length - 1
    weights(i) = 1.0 / weights.Length
Next

Learn.Creation = New ModelConstructor(Of DecisionStump)(x=>tree.Compute(x))
Dim Err As Double = Learn.Run(inputs, outputs, weights)

问题似乎出在这条线上：

Learn.Creation = New ModelConstructor(Of DecisionStump)(x=>tree.Compute(x))

如何在 Accord.Net 中使用 adaboost 或 boosting？如何调整我的代码以使其正常工作？我们将不胜感激。

Answer 1

这是一个较晚的响应，但对于那些将来可能会发现它有用的人来说，自 3.8.0 版以来，可以使用 Accord.NET Framework 学习增强决策树，如下所示：

// This example shows how to use AdaBoost to train more complex
// models than a simple DecisionStump. For example, we will use
// it to train a boosted Decision Trees.

// Let's use some synthetic data for that: The Yin-Yang dataset is 
// a simple 2D binary non-linear decision problem where the points 
// belong to each of the classes interwine in a Yin-Yang shape:
var dataset = new YinYang();
double[][] inputs = dataset.Instances;
int[] outputs = Classes.ToZeroOne(dataset.ClassLabels);

// Create an AdaBoost for Logistic Regression as:
var teacher = new AdaBoost<DecisionTree>()
{
    // Here we can specify how each regression should be learned:
    Learner = (param) => new C45Learning()
    {
        // i.e.
        // MaxHeight = 
        // MaxVariables = 
    },

    // Train until:
    MaxIterations = 50,
    Tolerance = 1e-5,
};

// Now, we can use the Learn method to learn a boosted classifier
Boost<DecisionTree> classifier = teacher.Learn(inputs, outputs);

// And we can test its performance using (error should be 0.11):
double error = ConfusionMatrix.Estimate(classifier, inputs, outputs).Error;

// And compute a decision for a single data point using:
bool y = classifier.Decide(inputs[0]); // result should false