Skip to content

Reduce memory footprint when using stochastic optimizers with shuffle activated #14075

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 19 commits into from
Jan 13, 2020
Merged

Reduce memory footprint when using stochastic optimizers with shuffle activated #14075

Changes from all commits
Commits
Show all changes
19 commits
Select commit Hold shift + click to select a range
fa2adcf
Reduce memory footprint when using stochastic optimizers with shuffle
meyer89 Jun 12, 2019
a575c72
Reduce the line length for flake8
meyer89 Jun 12, 2019
e66cfa6
Apply patch for fixed unit tests. Drop scipy.sparse.coo_matrix support
meyer89 Jun 17, 2019
85bbab5
Do not use safe_indexing due to low speed. Handle Pandas DataFrame in…
Jun 20, 2019
0400df2
Shorter comment for flake8
Jun 20, 2019
d53194e
Do not import unused function safe_indexing
Jun 20, 2019
df615a7
Use safe_indexing without take (should fix tests)
meyer89 Jun 22, 2019
2fa1535
Use slightly shorter comment for flake8
meyer89 Jun 22, 2019
62f662d
Remove not-required pandas Dataframe handling
meyer89 Jun 24, 2019
0c638d6
Merge remote-tracking branch 'github/master'
meyer89 Dec 16, 2019
dd2e796
Update import to avoid FutureWarning
meyer89 Dec 17, 2019
57f3d4a
Merge remote-tracking branch 'github/master'
meyer89 Dec 17, 2019
f148d93
Merge remote-tracking branch 'github/master'
meyer89 Dec 19, 2019
fbb2a64
Merge remote-tracking branch 'github/master'
meyer89 Jan 2, 2020
516958f
Apply suggestions from code review
meyer89 Jan 13, 2020
8235d4b
Merge remote-tracking branch 'origin/master'
meyer89 Jan 13, 2020
05f1f55
Reduce line length for flake8
meyer89 Jan 13, 2020
3872d4e
Remove whitespace in empty line (flake8)
meyer89 Jan 13, 2020
76d31bf
Merge remote-tracking branch 'github/master'
meyer89 Jan 13, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 20 additions & 6 deletions sklearn/neural_network/_multilayer_perceptron.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@
from ..preprocessing import LabelBinarizer
from ..utils import gen_batches, check_random_state
from ..utils import shuffle
from ..utils import _safe_indexing
from ..utils import check_array, check_X_y, column_or_1d
from ..exceptions import ConvergenceWarning
from ..utils.extmath import safe_sparse_dot
Expand Down Expand Up @@ -503,6 +504,7 @@ def _fit_stochastic(self, X, y, activations, deltas, coef_grads,
y_val = None

n_samples = X.shape[0]
sample_idx = np.arange(n_samples, dtype=int)

if self.batch_size == 'auto':
batch_size = min(200, n_samples)
Expand All @@ -512,12 +514,24 @@ def _fit_stochastic(self, X, y, activations, deltas, coef_grads,
try:
for it in range(self.max_iter):
if self.shuffle:
X, y = shuffle(X, y, random_state=self._random_state)
# Only shuffle the sample indices instead of X and y to
# reduce the memory footprint. These indices will be used
# to slice the X and y.
sample_idx = shuffle(sample_idx,
random_state=self._random_state)

accumulated_loss = 0.0
for batch_slice in gen_batches(n_samples, batch_size):
activations[0] = X[batch_slice]
if self.shuffle:
X_batch = _safe_indexing(X, sample_idx[batch_slice])
y_batch = y[sample_idx[batch_slice]]
else:
X_batch = X[batch_slice]
y_batch = y[batch_slice]

activations[0] = X_batch
batch_loss, coef_grads, intercept_grads = self._backprop(
X[batch_slice], y[batch_slice], activations, deltas,
X_batch, y_batch, activations, deltas,
coef_grads, intercept_grads)
accumulated_loss += batch_loss * (batch_slice.stop -
batch_slice.start)
Expand Down Expand Up @@ -664,7 +678,7 @@ def _predict(self, X):
y_pred : ndarray of shape (n_samples,) or (n_samples, n_outputs)
The decision function of the samples for each class in the model.
"""
X = check_array(X, accept_sparse=['csr', 'csc', 'coo'])
X = check_array(X, accept_sparse=['csr', 'csc'])

# Make sure self.hidden_layer_sizes is a list
hidden_layer_sizes = self.hidden_layer_sizes
Expand Down Expand Up @@ -928,7 +942,7 @@ def __init__(self, hidden_layer_sizes=(100,), activation="relu",
n_iter_no_change=n_iter_no_change, max_fun=max_fun)

def _validate_input(self, X, y, incremental):
X, y = check_X_y(X, y, accept_sparse=['csr', 'csc', 'coo'],
X, y = check_X_y(X, y, accept_sparse=['csr', 'csc'],
multi_output=True)
if y.ndim == 2 and y.shape[1] == 1:
y = column_or_1d(y, warn=True)
Expand Down Expand Up @@ -1336,7 +1350,7 @@ def predict(self, X):
return y_pred

def _validate_input(self, X, y, incremental):
X, y = check_X_y(X, y, accept_sparse=['csr', 'csc', 'coo'],
X, y = check_X_y(X, y, accept_sparse=['csr', 'csc'],
multi_output=True, y_numeric=True)
if y.ndim == 2 and y.shape[1] == 1:
y = column_or_1d(y, warn=True)
Expand Down