Source code for datasets.corrupted_mnist

# Copyright 2011 Hugo Larochelle. All rights reserved.
# 
# Redistribution and use in source and binary forms, with or without modification, are
# permitted provided that the following conditions are met:
# 
#    1. Redistributions of source code must retain the above copyright notice, this list of
#       conditions and the following disclaimer.
# 
#    2. Redistributions in binary form must reproduce the above copyright notice, this list
#       of conditions and the following disclaimer in the documentation and/or other materials
#       provided with the distribution.
# 
# THIS SOFTWARE IS PROVIDED BY Hugo Larochelle ``AS IS'' AND ANY EXPRESS OR IMPLIED
# WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
# FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL Hugo Larochelle OR
# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
# CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
# ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
# NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
# ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
# 
# The views and conclusions contained in the software and documentation are those of the
# authors and should not be interpreted as representing official policies, either expressed
# or implied, of Hugo Larochelle.

"""
Module ``datasets.corrupted_mnist`` gives access to the corrupted MNIST dataset.

This is a multi-label classification dataset, where the task is to
reconstruct the original binary image of an MNIST digit from a corrupted
version of it. The corruption was generated by randomly flipping a subset
of the pixels in the original image.

The original dataset from http://yann.lecun.com/exdb/mnist/ has been
preprocessed so that the inputs are binary. Corrupted MNIST was
generated by Volodymyr Mnih.


| **References:**
| The MNIST database of handwritten digits
| LeCun and Cortes
| http://yann.lecun.com/exdb/mnist/

"""

import mlpython.misc.io as mlio
import numpy as np
import os
from gzip import GzipFile as gfile

[docs]def load(dir_path,load_to_memory=False): """ Loads the corrupted MNIST dataset. The data is given by a dictionary mapping from strings ``'train'``, ``'valid'`` and ``'test'`` to the associated pair of data and metadata. The inputs and targets have been converted to a binary format. **Defined metadata:** * ``'input_size'`` * ``'target_size'`` * ``'length'`` """ input_size=784 target_size=784 dir_path = os.path.expanduser(dir_path) def load_line(line): tokens = line.split() return (np.array([int(i) for i in tokens[:input_size]]), np.array([int(i) for i in tokens[input_size:]])) train_file,valid_file,test_file = [os.path.join(dir_path, 'corrupted_mnist_' + ds + '.txt') for ds in ['train','valid','test']] # Get data train,valid,test = [mlio.load_from_file(f,load_line) for f in [train_file,valid_file,test_file]] lengths = [50000,10000,10000] if load_to_memory: train,valid,test = [mlio.MemoryDataset(d,[(input_size,),(target_size,)],[np.float64,np.float64],l) for d,l in zip([train,valid,test],lengths)] # Get metadata train_meta,valid_meta,test_meta = [{'input_size':input_size,'target_size':target_size, 'length':l} for l in lengths] return {'train':(train,train_meta),'valid':(valid,valid_meta),'test':(test,test_meta)}
[docs]def obtain(dir_path): """ Downloads the dataset to ``dir_path``. """ dir_path = os.path.expanduser(dir_path) print 'Downloading the dataset' import urllib urllib.urlretrieve('http://www.cs.toronto.edu/~larocheh/public/datasets/corrupted_mnist/mnist_corrupted_u.mat',os.path.join(dir_path,'mnist_corrupted_u.mat')) urllib.urlretrieve('http://www.cs.toronto.edu/~larocheh/public/datasets/corrupted_mnist/mnist_corrupted_v.mat',os.path.join(dir_path,'mnist_corrupted_v.mat')) urllib.urlretrieve('http://www.cs.toronto.edu/~larocheh/public/datasets/corrupted_mnist/mnist_corrupted_valid_u.mat',os.path.join(dir_path,'mnist_corrupted_valid_u.mat')) urllib.urlretrieve('http://www.cs.toronto.edu/~larocheh/public/datasets/corrupted_mnist/mnist_corrupted_valid_v.mat',os.path.join(dir_path,'mnist_corrupted_valid_v.mat')) urllib.urlretrieve('http://www.cs.toronto.edu/~larocheh/public/datasets/corrupted_mnist/mnist_corrupted_test_u.mat',os.path.join(dir_path,'mnist_corrupted_test_u.mat')) urllib.urlretrieve('http://www.cs.toronto.edu/~larocheh/public/datasets/corrupted_mnist/mnist_corrupted_test_v.mat',os.path.join(dir_path,'mnist_corrupted_test_v.mat')) # Writing everything into text files, to allow for not loading the data into memory def write_to_txt_file(u,v,filename): f = open(filename,'w') for u_t,v_t in zip(u,v): for i in range(len(u_t)): f.write(str(int(u_t[i]>127))+' ') for i in range(len(v_t)-1): f.write(str(int(v_t[i]>127))+' ') f.write(str(int(v_t[-1]>127))+'\n') f.close() import scipy.io u = scipy.io.loadmat(os.path.join(dir_path,'mnist_corrupted_u.mat'))['dat'] v = scipy.io.loadmat(os.path.join(dir_path,'mnist_corrupted_v.mat'))['dat'] write_to_txt_file(u,v,os.path.join(dir_path,'corrupted_mnist_train.txt')) u = scipy.io.loadmat(os.path.join(dir_path,'mnist_corrupted_valid_u.mat'))['dat'] v = scipy.io.loadmat(os.path.join(dir_path,'mnist_corrupted_valid_v.mat'))['dat'] write_to_txt_file(u,v,os.path.join(dir_path,'corrupted_mnist_valid.txt')) u = scipy.io.loadmat(os.path.join(dir_path,'mnist_corrupted_test_u.mat'))['dat'] v = scipy.io.loadmat(os.path.join(dir_path,'mnist_corrupted_test_v.mat'))['dat'] write_to_txt_file(u,v,os.path.join(dir_path,'corrupted_mnist_test.txt')) print 'Done '