NeuralNetwork/training/mnistreader.go

/*
 * MIT License
 *
 * Copyright (c) 2019 Alexey Edelev <semlanik@gmail.com>
 *
 * This file is part of NeuralNetwork project https://git.semlanik.org/semlanik/NeuralNetwork
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy of this
 * software and associated documentation files (the "Software"), to deal in the Software
 * without restriction, including without limitation the rights to use, copy, modify,
 * merge, publish, distribute, sublicense, and/or sell copies of the Software, and
 * to permit persons to whom the Software is furnished to do so, subject to the following
 * conditions:
 *
 * The above copyright notice and this permission notice shall be included in all copies
 * or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,
 * INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR
 * PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE
 * FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
 * DEALINGS IN THE SOFTWARE.
 */

package training

import (
	"encoding/binary"
	"io"
	"log"
	"os"

	mat "gonum.org/v1/gonum/mat"
)

type mnistReader struct {
	dataFilename              string
	resultsFilename           string
	validatorFilename         string
	validatorResultsFilename  string
	dataCount                 int
	validatorCount            int
	imageSize                 int
	buffered                  *mat.Dense
	resultsBuffered           *mat.Dense
	bufferedValidation        *mat.Dense
	resultsBufferedValidation *mat.Dense
}

func NewMNISTReader(dataFilename string, resultsFilename string, validatorFilename string, validatorResultsFilename string) (r *mnistReader) {
	r = &mnistReader{}

	r.dataCount, r.imageSize = openFileSet(dataFilename, resultsFilename)
	r.validatorCount, _ = openFileSet(validatorFilename, validatorResultsFilename)
	if r.dataCount <= 0 || r.imageSize <= 0 || r.validatorCount <= 0 {
		return nil
	}
	return
}

func (r *mnistReader) GetData(i int) (*mat.Dense, *mat.Dense) {
	if r.dataCount <= i {
		return nil, nil
	}

	return r.readData(r.dataFilename, r.resultsFilename, i)
}

func (r *mnistReader) DataCount() int {
	return r.dataCount
}

func (r *mnistReader) GetValidator(i int) (data *mat.Dense, result *mat.Dense) {
	if r.validatorCount <= i {
		return nil, nil
	}

	return r.readData(r.validatorFilename, r.validatorResultsFilename, i)
}

func (r *mnistReader) ValidatorCount() int {
	return r.validatorCount
}

func (r *mnistReader) readData(data string, result string, i int) (buffered, resultsBuffered *mat.Dense) {
	file, err := os.Open(data)
	if err != nil {
		return nil, nil
	}
	defer file.Close()

	resultsFile, err := os.Open(result)
	if err != nil {
		return nil, nil
	}
	defer resultsFile.Close()

	file.Seek(16+int64(r.imageSize*i), 0)
	resultsFile.Seek(8+int64(i), 0)

	buffer := make([]byte, r.imageSize)
	_, err = file.Read(buffer)

	if err == io.EOF {
		return nil, nil
	} else if err != nil {
		log.Fatal("File read error\n")
	}

	values := make([]float64, r.imageSize)
	for i, v := range buffer {
		values[i] = float64(v) / 255.0
	}

	buffered = mat.NewDense(r.imageSize, 1, values)

	buffer = make([]byte, 1)
	_, err = resultsFile.Read(buffer)
	if err != nil {
		log.Fatal("Result file read error\n")
	}

	num := int(buffer[0])

	resultsBuffered = mat.NewDense(10, 1, nil)
	resultsBuffered.Set(num, 0, 1.0)

	return buffered, resultsBuffered
}

func openFileSet(dataFilename string, resultsFilename string) (count int, imageSize int) {
	var err error
	data, err := os.Open(dataFilename)
	if err != nil {
		return -1, -1
	}
	defer data.Close()

	result, err := os.Open(resultsFilename)
	if err != nil {
		return -1, -1
	}
	defer result.Close()

	buffer := make([]byte, 16)
	data.Read(buffer)
	header := binary.BigEndian.Uint32(buffer[:4])
	if header != 0x00000803 {
		return -1, -1
	}
	count = int(binary.BigEndian.Uint32(buffer[4:8]))
	imageSize = int(binary.BigEndian.Uint32(buffer[8:12])) * int(binary.BigEndian.Uint32(buffer[12:16]))

	buffer = make([]byte, 8)
	result.Read(buffer)
	header = binary.BigEndian.Uint32(buffer[0:4])
	if header != 0x00000801 {
		return -1, -1
	}
	resultsCount := int(binary.BigEndian.Uint32(buffer[4:8]))
	if resultsCount != count {
		return -1, -1
	}

	return
}