iterator and concatenate_two_sets that can deal with per-time-point y

braindecode/__init__.py

@@ -0,0 +1 @@
+from braindecode.version import __version__

braindecode/datautil/iterators.py

@@ -218,7 +218,7 @@ def _yield_block_batches(self, X, y, start_stop_blocks_per_trial, shuffle):
         for i_blocks in blocks_per_batch:
             start_stop_blocks = i_trial_start_stop_block[i_blocks]
             batch = _create_batch_from_i_trial_start_stop_blocks(
-                X, y, start_stop_blocks)
+                X, y, start_stop_blocks, self.n_preds_per_input)
             yield batch
 
 
@@ -231,7 +231,7 @@ def _compute_start_stop_block_inds(i_trial_starts, i_trial_stops,
     ----------
     i_trial_starts: 1darray/list of int
         Indices of first samples to predict(!).
-    i_trial_stops: 1daray/list of int
+    i_trial_stops: 1darray/list of int
         Indices one past last sample to predict.
     input_time_length: int
     n_preds_per_input: int
@@ -299,12 +299,17 @@ def _get_start_stop_blocks_for_trial(i_trial_start, i_trial_stop,
     return start_stop_blocks
 
 
-def _create_batch_from_i_trial_start_stop_blocks(X, y, i_trial_start_stop_block):
+def _create_batch_from_i_trial_start_stop_blocks(X, y, i_trial_start_stop_block,
+                                                 n_preds_per_input=None):
     Xs = []
     ys = []
     for i_trial, start, stop in i_trial_start_stop_block:
         Xs.append(X[i_trial][:,start:stop])
-        ys.append(y[i_trial])
+        if not hasattr(y[i_trial], '__len__'):
+            ys.append(y[i_trial])
+        else:
+            assert n_preds_per_input is not None
+            ys.append(y[i_trial][stop-n_preds_per_input:stop])
     batch_X = np.array(Xs)
     batch_y = np.array(ys)
     # add empty fourth dimension if necessary

braindecode/datautil/splitters.py

@@ -33,17 +33,22 @@ def concatenate_two_sets(set_a, set_b):
     -------
     concatenated_set: :class:`.SignalAndTarget`
     """
-    if hasattr(set_a.X, 'ndim') and hasattr(set_b.X, 'ndim'):
-        new_X = np.concatenate((set_a.X, set_b.X), axis=0)
-    else:
-        if hasattr(set_a.X, 'ndim'):
-            set_a.X = set_a.X.tolist()
-        if hasattr(set_b.X, 'ndim'):
-            set_b.X = set_b.X.tolist()
-        new_X = set_a.X + set_b.X
-    new_y = np.concatenate((set_a.y, set_b.y), axis=0)
+    new_X = concatenate_np_array_or_add_lists(set_a.X, set_b.X)
+    new_y = concatenate_np_array_or_add_lists(set_a.y, set_b.y)
     return SignalAndTarget(new_X, new_y)
 
+def concatenate_np_array_or_add_lists(a, b):
+    if hasattr(a, 'ndim') and hasattr(b, 'ndim'):
+        new = np.concatenate((a, b), axis=0)
+    else:
+        if hasattr(a, 'ndim'):
+            a = a.tolist()
+        if hasattr(b, 'ndim'):
+            b = b.tolist()
+        new = a + b
+    return new
+
+
 
 def split_into_two_sets(dataset, first_set_fraction=None, n_first_set=None):
     """

braindecode/torch_ext/losses.py

@@ -1,9 +1,10 @@
 import torch as th
 
 
-def log_categorical_crossentropy(logpreds, targets, dims=None):
+def log_categorical_crossentropy_1_hot(logpreds, targets, dims=None):
     """
-    Returns log categorical crossentropy for given log-predictions and targets.
+    Returns log categorical crossentropy for given log-predictions and targets,
+    targets should be one-hot-encoded.
     
     Computes :math:`-\mathrm{logpreds} \cdot \mathrm{targets}`
     
@@ -12,6 +13,7 @@ def log_categorical_crossentropy(logpreds, targets, dims=None):
     logpreds: `torch.autograd.Variable`
         Logarithm of softmax output.
     targets: `torch.autograd.Variable`
+        One-hot encoded targets
     dims: int or iterable of int, optional.
         Compute sum across these dims
 
@@ -31,6 +33,39 @@ def log_categorical_crossentropy(logpreds, targets, dims=None):
     return result
 
 
+def log_categorical_crossentropy(log_preds, targets):
+    """
+    Returns log categorical crossentropy for given log-predictions and targets.
+    
+    Computes :math:`-\mathrm{logpreds} \cdot \mathrm{targets}` if you assume
+    targets to be one-hot-encoded. Also works for targets that are not
+    one-hot-encoded, in this case only uses targets that are in the range
+    of the expected class labels, i.e., [0,log_preds.size()[1]-1].
+
+    Parameters
+    ----------
+    log_preds: torch.autograd.Variable`
+        Logarithm of softmax output.
+    targets: `torch.autograd.Variable`
+
+    Returns
+    -------
+
+    loss: `torch.autograd.Variable`
+    """
+    if log_preds.size() == targets.size():
+        return log_categorical_crossentropy_1_hot(log_preds, targets)
+    n_classes = log_preds.size()[1]
+    n_elements = 0
+    losses = []
+    for i_class in range(n_classes):
+        mask = targets == i_class
+        mask = mask.type_as(log_preds)
+        n_elements -= th.sum(mask)
+        losses.append(th.sum(mask * log_preds[:,i_class]))
+    return th.sum(th.stack(losses)) / n_elements
+
+
 def l2_loss(model):
     losses = [th.sum(p * p) for p in model.parameters()]
     return sum(losses)

braindecode/version.py

@@ -0,0 +1 @@
+__version__ = "0.2.0"

docs/Makefile

@@ -21,3 +21,13 @@ apidoc:
 # "make mode" option.  $(O) is meant as a shortcut for $(SPHINXOPTS).
 %: Makefile
 	@$(SPHINXBUILD) -M $@ "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)
+
+
+removeipynbcheckpoints: Makefile
+	rm -rf notebooks/.ipynb_checkpoints/ notebooks/visualization/.ipynb_checkpoints/
+
+removesource: Makefile
+	rm -rf source/
+
+rmanddoc: removesource removeipynbcheckpoints apidoc html
+	echo "Done"

docs/conf.py

@@ -82,8 +82,9 @@ def find_source():
 ## Default flags used by autodoc directives
 autodoc_default_flags = ['members', 'show-inheritance']
 
-exclude_patterns = ['_build', '_templates',]
-
+exclude_patterns = ['_build', '_templates']
+
+
 napoleon_google_docstring = False
 napoleon_use_param = False
 napoleon_use_ivar = True
@@ -110,9 +111,10 @@ def find_source():
 # built documents.
 #
 # The short X.Y version.
-version = '0.1.9'
+import braindecode
+version = braindecode.__version__
 # The full version, including alpha/beta/rc tags.
-release = '0.1.9'
+release = version
 
 # The language for content autogenerated by Sphinx. Refer to documentation
 # for a list of supported languages.
@@ -211,20 +213,3 @@ def find_source():
      author, 'Braindecode', 'One line description of project.',
      'Miscellaneous'),
 ]
-
-
-## mock stuff
-"""
-import sys
-from mock import Mock as MagicMock
-#from unittest.mock import MagicMock
-
-class Mock(MagicMock):
-    @classmethod
-    def __getattr__(cls, name):
-            return MagicMock()
-
-MOCK_MODULES = ['torch', 'h5py',]
-sys.modules.update((mod_name, Mock()) for mod_name in MOCK_MODULES)
-
-"""

docs/index.rst

@@ -3,7 +3,7 @@ Welcome to Braindecode
 
 A deep learning toolbox to decode raw time-domain EEG.
 
-For EEG researchers that want to want to work with deep learning and
+For EEG researchers that want to work with deep learning and
 deep learning researchers that want to work with EEG data.
 For now focussed on convolutional networks.
 

docs/notebooks/Cropped_Decoding.ipynb

@@ -449,6 +449,9 @@
   }
  ],
  "metadata": {
+  "git": {
+   "keep_outputs": true
+  },
   "kernelspec": {
    "display_name": "Python 3",
    "language": "python",

docs/source/braindecode.rst

@@ -30,4 +30,12 @@ braindecode\.util module
     :undoc-members:
     :show-inheritance:
 
+braindecode\.version module
+---------------------------
+
+.. automodule:: braindecode.version
+    :members:
+    :undoc-members:
+    :show-inheritance:
+
 

setup.py

@@ -9,13 +9,16 @@
     long_description = f.read()
 
 
+# This will add __version__ to version dict
+version = {}
+with open(path.join(here, 'braindecode/version.py'), encoding='utf-8') as (
+        version_file):
+    exec(version_file.read(), version)
+
 setup(
     name='Braindecode',
 
-    # Versions should comply with PEP440.  For a discussion on single-sourcing
-    # the version across setup.py and the project code, see
-    # http://packaging.python.org/en/latest/tutorial.html#version
-    version='0.1.9', # TODO: read from __init__.py?
+    version=version['__version__'],
 
     description='A deep learning toolbox to decode raw time-domain EEG.',
     long_description=long_description,