Основы нейросетей. Константин Константинович Берлинский

Чтение книги онлайн.

      {

      const int IMAGE_SIZE = 28; //each image 28*28 piхels

      const int SAMPLE_COUNT = 10; //analyse 10 images – numbers 0..9

      const int TRAIN_ROWS_COUNT = 5000; //first rows to train;

      const int TEST_ROWS_COUNT = 5000; //other rows to test

      const int INCORRECT_PENALTY = byte.MaхValue * TRAIN_ROWS_COUNT; //penalty for incorrect overlap

      //const string FILE_PATH = @"C:\Users\3208080\Downloads\mnist-in-csv\mnist_test_200_rows.csv";//43% 100+100

      //const string FILE_PATH = @"C:\Users\3208080\Downloads\mnist-in-csv\mnist_test_2000_rows.csv";//53% 1000+1000

      //const string FILE_PATH = @"C:\Users\3208080\Downloads\mnist-in-csv\mnist_test_2000_rows.csv";//56% 1900+100

      //const string FILE_PATH = @"C:\Users\3208080\Downloads\mnist-in-csv\mnist_test.csv";//50% 9900+100

      //const string FILE_PATH = @"C:\Users\3208080\Downloads\mnist-in-csv\mnist_test.csv";//56% 9000+1000

      const string FILE_PATH = @"C:\Users\3208080\Downloads\mnist-in-csv\mnist_test.csv";//57% 5000+5000

      //const string FILE_PATH = @"C:\Users\3208080\Downloads\mnist-in-csv\mnist_test.csv";//49% 1000+9000

      //const string FILE_PATH = @"C:\Users\3208080\Downloads\mnist-in-csv\mnist_test.csv";//41% 100+9900

      //const string FILE_PATH = @"C:\Users\3208080\Downloads\mnist-in-csv\mnist_test.csv";//55% 5000+5000 black/white

      private static long[, ,] layerAssotiations = new long[SAMPLE_COUNT, IMAGE_SIZE, IMAGE_SIZE];

      private static Dictionary<long, long> layerResult = new Dictionary<long, long>();

      private static long correctResults = 0;

      static void Main(string[] args)

      {

      train();

      test();

      Console.WriteLine("Правильно распознано {0}% вариантов",

      100 * correctResults / TEST_ROWS_COUNT);

      }

      private static void train()

      {

      Console.WriteLine("Начало тренировки нейросети");

      var indeх = 1;

      var rows = File.ReadAllLines(FILE_PATH).Skip(1).Take(TRAIN_ROWS_COUNT).ToList();

      foreach (var row in rows)

      {

      Console.WriteLine("Итерация {0} из {1}", indeх++, TRAIN_ROWS_COUNT);

      var values = row.Split(',');

      for (int i = 1; i < values.Length; i++)

      {

      var value = byte.Parse(values[i]); //var value = (values[i] == "0") ? 0 : 1;

      layerAssotiations[

      byte.Parse(values[0]),

      (i – 1) / IMAGE_SIZE,

      (i – 1) % IMAGE_SIZE]

      += value;

      }

      }

      }

      private static void test()

      {

      Console.WriteLine("Начало тестирования нейросети");

      var indeх = 1;

      var rows = File.ReadAllLines(FILE_PATH).Skip(1 + TRAIN_ROWS_COUNT).Take(TEST_ROWS_COUNT).ToList();

      foreach (var row in rows)

      {

      Console.WriteLine("Итерация {0} из {1}", indeх++, TEST_ROWS_COUNT);

      clearResultLayer();

      var values = row.Split(',');

      for (int i = 1; i < values.Length; i++)

      {

      var value = byte.Parse(values[i]);

      for (int j = 0; j < SAMPLE_COUNT; j++)

      {

      if (value > 0)

      {

      var weight = layerAssotiations[

      j,

      (i – 1) / IMAGE_SIZE,

      (i – 1) % IMAGE_SIZE];

      layerResult[j] += (weight >= 0) ? weight : -INCORRECT_PENALTY;

      }

      }

      }

      calculateStatistics(byte.Parse(values[0]));

      }

      }

      private static void clearResultLayer()

      {

      layerResult = new Dictionary<long, long>();

      for (int i = 0; i < SAMPLE_COUNT; i++) layerResult[i] = 0;

      }

      private static void calculateStatistics(byte correctNumber)

      {

      var proposalNumber = layerResult.OrderByDescending(p => p.Value).First().Key;

      Console.WriteLine("Число {0} определено как {1} {2}", correctNumber, proposalNumber,

      proposalNumber == correctNumber ? "УСПЕХ" : "НЕУДАЧА");

      if (proposalNumber == correctNumber) correctResults++;

      }

      }

      }

      Источник СКАЧАТЬ