//    $Revision: 1.1 $

using System;
using System.Collections.Generic;
using System.Text;
using fvalgcli;

namespace TC
{
    public partial class FIE
    {
        /// <summary>
        /// ニューラルネットワークの学習(確率的バックプロパゲーション法)
        /// </summary>
        /// <param name="hNeuro">ニューラルネットワークオブジェクトのハンドル</param>
        public void fnFIE_neuro_performance(FHANDLE hNeuro)
        {
            int status = (int)fvalgcli.f_err.F_ERR_NONE;

            DOUBLE_PTR vect_inp = DOUBLE_PTR.Zero;    // 入力ベクトル.
            DOUBLE_PTR vect_tch = DOUBLE_PTR.Zero;    // 教師ベクトル.

            // 入力ベクトル、教師ベクトル初期化用テーブル.
            double[,] data_tbl =
            {
                { 0.0, 0.0, 0.0 },
                { 0.0, 1.0, 1.0 },
                { 1.0, 0.0, 1.0 },
                { 1.0, 1.0, 0.0 }
            };

            try
            {
                // 入力ベクトル、教師ベクトルの初期化.
                vect_inp = DOUBLE_PTR.alloc(INPUT_NEURON * LEARN_COUNT);
                vect_tch = DOUBLE_PTR.alloc(OUTPUT_NEURON * LEARN_COUNT);

                // --- 入力ベクトル.
                for (int i = 0; i < LEARN_COUNT; i++)
                {
                    vect_inp[INPUT_NEURON * i + 0] = data_tbl[i % 4, 0];
                    vect_inp[INPUT_NEURON * i + 1] = data_tbl[i % 4, 1];
                }

                // --- 教師ベクトル.
                for (int i = 0; i < LEARN_COUNT; i++)
                {
                    vect_tch[OUTPUT_NEURON * i] = data_tbl[i % 4, 2];
                }

                // 学習.
                for (int i = 0; i < LEARN_COUNT; i++)
                {
                    DOUBLE_PTR _vect_inp = vect_inp.at(INPUT_NEURON * i);
                    DOUBLE_PTR _vect_tch = vect_tch.at(OUTPUT_NEURON * i);
                    double error = 0;    // 教師ベクトルとの誤差.

                    status = api.fnFIE_neuro_performance(hNeuro, _vect_inp, _vect_tch, ref error);

                    Assert.IsTrue(status == (int)f_err.F_ERR_NONE, "エラーが発生しました。({0})", (fvalgcli.f_err)status);

                    // 確認.
                    if (i < 10)
                        Console.WriteLine("fnFIE_neuro_performance[{0}] : error = {1}", i, error);
                }
            }
            finally
            {
                vect_inp.Dispose();
                vect_tch.Dispose();
            }
        }
    }
}

'    $Revision: 1.1 $

Imports System.Collections.Generic
Imports System.Text
Imports fvalgcli

Public Partial Class FIE
    ''' <summary>
    ''' ニューラルネットワークの学習(確率的バックプロパゲーション法)
    ''' </summary>
    ''' <param name="hNeuro">ニューラルネットワークオブジェクトのハンドル</param>
    Public Sub fnFIE_neuro_performance(hNeuro As FHANDLE)
        Dim status As Integer = CInt(fvalgcli.f_err.F_ERR_NONE)

        Dim vect_inp As DOUBLE_PTR = DOUBLE_PTR.Zero
        ' 入力ベクトル.
        Dim vect_tch As DOUBLE_PTR = DOUBLE_PTR.Zero
        ' 教師ベクトル.
        ' 入力ベクトル、教師ベクトル初期化用テーブル.
        Dim data_tbl As Double(,) = {{0.0, 0.0, 0.0}, {0.0, 1.0, 1.0}, {1.0, 0.0, 1.0}, {1.0, 1.0, 0.0}}

        Try
            ' 入力ベクトル、教師ベクトルの初期化.
            vect_inp = DOUBLE_PTR.alloc(INPUT_NEURON * LEARN_COUNT)
            vect_tch = DOUBLE_PTR.alloc(OUTPUT_NEURON * LEARN_COUNT)

            ' --- 入力ベクトル.
            For i As Integer = 0 To LEARN_COUNT - 1
                vect_inp(INPUT_NEURON * i + 0) = data_tbl(i Mod 4, 0)
                vect_inp(INPUT_NEURON * i + 1) = data_tbl(i Mod 4, 1)
            Next

            ' --- 教師ベクトル.
            For i As Integer = 0 To LEARN_COUNT - 1
                vect_tch(OUTPUT_NEURON * i) = data_tbl(i Mod 4, 2)
            Next

            ' 学習.
            For i As Integer = 0 To LEARN_COUNT - 1
                Dim _vect_inp As DOUBLE_PTR = vect_inp.at(INPUT_NEURON * i)
                Dim _vect_tch As DOUBLE_PTR = vect_tch.at(OUTPUT_NEURON * i)
                Dim [error] As Double = 0
                ' 教師ベクトルとの誤差.
                status = api.fnFIE_neuro_performance(hNeuro, _vect_inp, _vect_tch, [error])

                Assert.IsTrue(status = CInt(f_err.F_ERR_NONE), "エラーが発生しました。({0})", CType(status, fvalgcli.f_err))

                ' 確認.
                If i < 10 Then
                    Console.WriteLine("fnFIE_neuro_performance[{0}] : error = {1}", i, [error])
                End If
            Next
        Finally
            vect_inp.Dispose()
            vect_tch.Dispose()
        End Try
    End Sub
End Class