kelpnetの作法 その9

概要

kelpnetの作法を調べてみた。
九九やってみた。

結果

サンプルコード

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using KelpNet;

namespace ConsoleApp5
{
    class Program
    {
        const int EPOCH = 2001;
        const int N = 100;
        static void Main(string[] args)
        {
            Real[][] trainData = new Real[N][];
            Real[][] trainLabel = new Real[N][];
            int k;
            for (int i = 0; i < 10; i++)
            {
                for (int j = 0; j < 10; j++)
                {
                    k = i * 10 + j;
                    int a = (i >> 3) % 2;
                    int b = (i >> 2) % 2;
                    int c = (i >> 1) % 2;
                    int d = i % 2;
                    int e = (j >> 3) % 2;
                    int f = (j >> 2) % 2;
                    int g = (j >> 1) % 2;
                    int h = j % 2;
                    trainData[k] = new[] {
                        (Real)a,
                        (Real)b,
                        (Real)c,
                        (Real)d,
                        (Real)e,
                        (Real)f,
                        (Real)g,
                        (Real)h
                    };
                }
            }
            int l;
            for (int i = 0; i < 10; i++)
            {
                for (int j = 0; j < 10; j++)
                {
                    k = i * 10 + j;
                    l = i * j;
                    int a = (l >> 6) % 2;
                    int b = (l >> 5) % 2;
                    int c = (l >> 4) % 2;
                    int d = (l >> 3) % 2;
                    int e = (l >> 2) % 2;
                    int f = (l >> 1) % 2;
                    int g = l % 2;
                    trainLabel[k] = new[] {
                        (Real)a,
                        (Real)b,
                        (Real)c,
                        (Real)d,
                        (Real)e,
                        (Real)f,
                        (Real)g
                    };
                }
            }
            FunctionStack nn = new FunctionStack(new Linear(8, 40, name: "l1 Linear"), new TanhActivation(name: "act"), new Linear(40, 7, name: "l2 Linear"));
            nn.SetOptimizer(new MomentumSGD());
            Console.WriteLine("Train Start...");
            for (int i = 0; i < EPOCH; i++)
            {
                Real loss = 0;
                for (int j = 0; j < N; j++)
                {
                    loss += Trainer.Train(nn, trainData[j], trainLabel[j], new MeanSquaredError());
                }
                if (i % 100 == 0)
                {
                    Console.WriteLine("loss:" + loss / N);
                }
            }
            Console.WriteLine("Test Start...");
            Console.Write("   ");
            for (int i = 1; i < 10; i++)
            {
                Console.Write(String.Format("{0, 2} ", i));
            }
            Console.WriteLine("");
            for (int i = 1; i < 10; i++)
            {
                Console.Write(String.Format("{0, 2} ", i));
                for (int j = 1; j < 10; j++)
                {
                    k = i * 10 + j;
                    l = 0;
                    NdArray result = nn.Predict(trainData[k])[0];
                    if (result.Data[0] > 0.5)
                    {
                        l += 64;
                    }
                    if (result.Data[1] > 0.5)
                    {
                        l += 32;
                    }
                    if (result.Data[2] > 0.5)
                    {
                        l += 16;
                    }
                    if (result.Data[3] > 0.5)
                    {
                        l += 8;
                    }
                    if (result.Data[4] > 0.5)
                    {
                        l += 4;
                    }
                    if (result.Data[5] > 0.5)
                    {
                        l += 2;
                    }
                    if (result.Data[6] > 0.5)
                    {
                        l += 1;
                    }
                    Console.Write(String.Format("{0, 2} ", l));
                }
                Console.WriteLine("");
            }
            Console.WriteLine("Press any key to exit.");
            Console.ReadKey();
        }
    }
}

以上。