""" This module implements a fast and memory-efficient (no memory copying)
loader for the svmlight / libsvm sparse dataset format. """
# Authors: Mathieu Blondel <mathieu@mblondel.org>
# Lars Buitinck <L.J.Buitinck@uva.nl>
# License: Simple BSD.
from _svmlight_format import _load_svmlight_file, _dump_svmlight_file
import numpy as np
[docs]def load_svmlight_file(file_path, buffer_mb=40, query_id=False):
"""Load datasets in the svmlight / libsvm format into sparse CSR matrix
This format is a text-based format, with one sample per line. It does
not store zero valued features hence is suitable for sparse dataset.
The first element of each line can be used to store a target variable
to predict.
This format is used as the default format for both svmlight and the
libsvm command line programs.
Parsing a text based source can be expensive. When working on
repeatedly on the same dataset, it is recommended to wrap this
loader with joblib.Memory.cache to store a memmapped backup of the
CSR results of the first call and benefit from the near instantaneous
loading of memmapped structures for the subsequent calls.
Parameters
----------
file_path: str
Path to a file to load.
buffer_mb : integer
Buffer size to use for low level read
query_id : bool
True if the query ids has to be loaded, false otherwise
Returns
-------
(X, y, [query_ids])
where X is a dense numpy matrix of shape (n_samples, n_features) and type dtype,
y is a ndarray of shape (n_samples,).
query_ids is a ndarray of shape(nsamples,) if query_id is True, it is not returned otherwise
"""
data, labels, qids = _load_svmlight_file(file_path, buffer_mb)
# reshape the numpy array into a matrix
n_samples = len(labels)
n_features = len(data) / n_samples
data.shape = (n_samples, n_features)
if data.dtype != np.float32:
new_data = data.astype(dtype=np.float32)
del data
data = new_data
# Convert infinite values to max_float representation (SVM reader problem)
# This patch is needed because some dataset have infinite values and because
# the split condition is <=, while sole software uses <. In order to
# reconduct the former condition to the latter, we slightly decrease the
# split. However, slightly decreasing inf does not have any effect.
data[data == np.inf] = np.finfo(data.dtype).max
if not query_id:
return data, labels
else:
return data, labels, qids
[docs]def load_svmlight_files(files, buffer_mb=40, query_id=False):
"""Load dataset from multiple files in SVMlight format
This function is equivalent to mapping load_svmlight_file over a list of
files, except that the results are concatenated into a single, flat list
and the samples vectors are constrained to all have the same number of
features.
Parameters
----------
files : iterable over str
Paths to files to load.
n_features: int or None
The number of features to use. If None, it will be inferred from the
first file. This argument is useful to load several files that are
subsets of a bigger sliced dataset: each subset might not have
examples of every feature, hence the inferred shape might vary from
one slice to another.
Returns
-------
[X1, y1, ..., Xn, yn]
where each (Xi, yi, [comment_i, query_id_i]) tuple is the result from load_svmlight_file(files[i]).
Rationale
---------
When fitting a model to a matrix X_train and evaluating it against a
matrix X_test, it is essential that X_train and X_test have the same
number of features (X_train.shape[1] == X_test.shape[1]). This may not
be the case if you load them with load_svmlight_file separately.
See also
--------
load_svmlight_file
"""
files = iter(files)
result = list(load_svmlight_file(next(files), buffer_mb, query_id=query_id))
for f in files:
result += load_svmlight_file(f, buffer_mb, query_id=query_id)
return result
[docs]def dump_svmlight_file(X, y, f, query_id=None, zero_based=True):
"""Dump the dataset in svmlight / libsvm file format.
This format is a text-based format, with one sample per line. It does
not store zero valued features hence is suitable for sparse dataset.
The first element of each line can be used to store a target variable
to predict.
Parameters
----------
X : CSR sparse matrix, shape = [n_samples, n_features]
Training vectors, where n_samples is the number of samples and
n_features is the number of features.
y : array-like, shape = [n_samples]
Target values.
f : str
Specifies the path that will contain the data.
comment : list, optional
Comments to append to each row after a # character
If specified, len(comment) must equal n_samples
query_id: list, optional
Query identifiers to prepend to each row
If specified, len(query_id) must equal n_samples
zero_based : boolean, optional
Whether column indices should be written zero-based (True) or one-based
(False).
"""
if hasattr(f, "write"):
raise ValueError("File handler not supported. Use a file path.")
# todo: this test is ok if we know the number of features! to divide X.shape[0] with
# if X.shape[0] != y.shape[0]:
# raise ValueError("X.shape[0] and y.shape[0] should be the same, "
# "got: %r and %r instead." % (X.shape[0], y.shape[0]))
if query_id is None:
query_id = []
# elif X.shape[0] != len(query_id):
# raise ValueError("X.shape[0] and len(query_id) should be the same, "
# "got: %r and %r instead." % (X.shape[0], len(query_id)))
X = np.array(X, dtype=np.float32)
y = np.array(y, dtype=np.float32)
_dump_svmlight_file(f, X, y, query_id, int(zero_based))