How to use ActivationNetwork of Accord.NET properly to do machine learning tasks such as BackPropagationLearning

[FurkanGozukara]

Source of the question : https://stackoverflow.com/questions/64827817/how-to-use-activationnetwork-of-accord-net-properly-to-do-machine-learning-tasks

It is sad that the examples of Accord.NET is extremely primitive and insufficient. Therefore, I am having hard time to figure out how to use the system properly.

The entire source code of my application is uploaded to here : https://github.com/FurkanGozukara/CSE419-Artificial-Intelligence-and-Machine-Learning-2020/tree/master/source%20codes/lecture%206%20perceptron%20example

The video of this lecture is here (3 hours 11 minutes) : https://youtu.be/qrklFBewlJA

My biggest question is about how to provide output classes?

Lets say for the dataset abalone (http://archive.ics.uci.edu/ml/datasets/Abalone) there are 28 output classes which are the age of abalone : https://github.com/FurkanGozukara/CSE419-Artificial-Intelligence-and-Machine-Learning-2020/blob/master/source%20codes/lecture%206%20perceptron%20example/lecture%206%20perceptron%20example/bin/Debug/netcoreapp3.1/abalone.data

How to compose the output class?

Like this way?

            int irNumberOfExamples = File.ReadAllLines(srFileName).Count();

        double[][] input = new double[irNumberOfExamples][];
        double[][] output = new double[irNumberOfExamples][];

        List<double> lstOutPutClasses = new List<double>();

        NumberFormatInfo formatProvider = new NumberFormatInfo();
        formatProvider.NumberDecimalSeparator = ".";
        formatProvider.NumberGroupSeparator = ",";

        foreach (var vrPerLine in File.ReadAllLines(srFileName))
        {
            var vrOutPut = Convert.ToDouble(vrPerLine.Split(',').Last(), formatProvider);

            if (lstOutPutClasses.Contains(vrOutPut) == false)
            {
                lstOutPutClasses.Add(vrOutPut);
            }
        }

        int irCounter = 0;
        foreach (var vrPerLine in File.ReadAllLines(srFileName))
        {
            input[irCounter] = vrPerLine.Split(',').SkipLast(1).
                Select(pr => Convert.ToDouble(pr.Replace("I", "0.0").Replace("M", "0.5").Replace("F", "1.0"), formatProvider)).ToArray();

            var vrCurrentOutClass = Convert.ToDouble(vrPerLine.Split(',').Last(), formatProvider);

            output[irCounter][lstOutPutClasses.IndexOf(vrCurrentOutClass)] = 1;

            irCounter++;
        }

This generates the output class like below

And here the code of training part

   int irFinalClassCount = lstOutPutClasses.Count;

        double learningRate = 0.1;
        int irNumberOfFeatures = input[0].Length;

        ActivationNetwork network3 = new ActivationNetwork(
      new SigmoidFunction(2),//activation function
       irNumberOfFeatures,//input layer equal number of features
       12,// 12 neurons at the hidden layer_1
       irFinalClassCount); //output layer equal to number of output classes
                           
        BackPropagationLearning bpteacher = new BackPropagationLearning(network3);
        bpteacher.LearningRate = 0.1;
        bpteacher.Momentum = 0.5;

        for (int i = 0; i < 200000; i++)
        {
            double error = bpteacher.RunEpoch(input, output);//train the algorithm

            var vrAcc = calculateAcurracy(network3, input, output);

            Console.WriteLine("BackPropagationLearning -> " + i + ", Error = " + error.ToString("N2") + "\t\t accuracy: " + vrAcc);
        }

So my second question is about calculating accuracy. The BackPropagationLearning algorithm generates weights for each output neuron. So I counted the highest one as the prediction. Is my approach and code correct?

I am believe I am on the right track but I want to be sure by getting feedback of an Accord.NET library expert