morphoclass.utils module

Miscellaneous utilities.

morphoclass.utils.add_metadata(checkpoint)

Add metadata information to the checkpoint.

The metadata information includes - A timestamp in ISO format - The name of the package (should be “morphoclass”) - The version of the package - Installed packages

Parameters

checkpoint (dict) – The checkpoint dictionary.

morphoclass.utils.are_you_sure()

Interactive prompt asking “Are you sure? (y/N)”.

Returns

True if the user answered with “y”, otherwise False.

Return type

bool

morphoclass.utils.convert_morphologies(input_dir, old_ext, new_ext, notebook=False)

Convert a collection of morphologies to a new format.

All files matching the old extension will be read recursively from the input directory and its subdirectories. A new output directory will then be created by appending and underscore and the new extension to the input directory name, e.g. the directory “my_files” will become “my_files_swc” if the new extension is “swc”.

Note that there are no checks if the output directory already exists and files might be overwritten in the case that it does!

Parameters

  • input_dir (str or pathlib.Path) – The directory containing the input morphologies. All subdirectories will be processed recursively.

  • old_ext (str) – The file extension to convert from. Only files with this extension will be considered.

  • new_ext (str) – The new extension to convert to. Must be a file extension supported by MorphIO.

  • notebook (bool) – If true then a widget progress bar will be displayed instead of an ASCII one.

morphoclass.utils.dict2kwargs(d)

Convert method parameters dict to keyword string for the HTML report.

Parameters

d (dict, str) – Dictionary to convert or empty string if no parameters are available.

Returns

String of dictionary shown as keywords.

Return type

str

morphoclass.utils.find_point_on_branch(tree: Tree, start: int, end: int) → list[int]

Find all the points in a branch specified by its start and end points.

Parameters

  • tree – A TMD tree.

  • start – Index of the starting point of the branch

  • end – Index of the ending point of the branch

Returns

A list of indices of points on the given branch. The indices refer to the non-simplified tree.

Return type

list

morphoclass.utils.from_morphio(root_section)

Convert a MorphIO root section to a TMD tree.

Parameters

root_section – A MorphIO root section.

Returns

tree – A TMD tree.

Return type

tmd.Tree.Tree.Tree

morphoclass.utils.from_morphio_to_tmd(morphio_neuron, remove_duplicates=False)

Change a neuron in MorphIO format to TMD neuron format.

Parameters

  • morphio_neuron – A MorphIO neuron.

  • remove_duplicates (bool) – Whether or not to remove duplicate points at section joints.

Returns

A TMD neuron.

Return type

tmd_neuron

morphoclass.utils.from_tmd_to_morphio(tmd_neuron)

Change a neuron in TMD format to MorphIO neuron format.

Parameters

tmd_neuron – A TMD neuron.

Returns

Return type

A MorphIO neuron.

morphoclass.utils.get_loader(samples, idx, batch_size=None)

Create a DataLoader from a list of samples and a given index.

Parameters

  • samples – A list of samples.

  • idx – A list of indices that specifies which samples should be included.

  • batch_size (int, optional) – Number of training samples per batch. By default a full batch will be used.

Returns

A DataLoader object.

Return type

torch_geometric.data.DataLoader

morphoclass.utils.get_stratified_split(dataset, val_size=None, random_state=None)

Get indices for a stratified train/val split.

Parameters

  • dataset (morphoclass.data.MorphologyDataset) – The dataset to split.

  • val_size (flaot) – The relative size of the validation set.

  • random_state – A random state for the StratifiedShuffleSplit class.

Returns

  • train_idx (list) – The indices of the training set.

  • val_idx – The indices of the validation set.

morphoclass.utils.load_var(file_name)

Load a serialized variable from the given file.

Parameters

file_name (str) – The file from which to load the variable.

morphoclass.utils.make_torch_deterministic(make_torch_traceback_nicer=True)

Make GPU computations using torch deterministic.

Note that in some cases this may have a performance impact. For further information see https://pytorch.org/docs/stable/notes/randomness.html

Parameters

make_torch_traceback_nicer (bool, optional) – make CUDA generated error tracebacks (“device-side assert triggered”) more explicit. Useful for debugging.

morphoclass.utils.morphio_root_section_to_tmd_tree(root_section, remove_duplicates=True)

Create a TMD tree class from a MorphIO root section.

Parameters

  • root_section – A MorphIO root section

  • remove_duplicates (bool) – If true then duplicate points at section joins are removed.

Returns

tree – A TMD tree.

Return type

tmd.Tree.Tree.Tree

morphoclass.utils.no_print(function)

Decorate a function to suppress all output to stdout.

morphoclass.utils.normalize_features(samples, max_values=None)

Normalize all sample features by their maximum.

Parameters

  • samples (iterable) – A list of samples to be normalized.

  • max_values (float or torch.Tensor) – A list of the max values features_wise.

Returns

  • samples – The normalized samples.

  • max_values – The max values that were used for normalization.

morphoclass.utils.np_temp_seed(seed)

Contextmanager for setting a temporary numpy seed.

Parameters

seed (int) – The temporary numpy seed.

morphoclass.utils.print_error(msg)

Print an error message.

morphoclass.utils.print_message_to_stderr(header, msg)

Print a message to stderr.

Parameters

  • header (str) – A header to prepend to the message.

  • msg (str) – The message.

morphoclass.utils.print_warning(msg)

Print a warning message.

morphoclass.utils.read_apical_from_file(path, label, nodes_features=None, inverted_apical=True)

Read the neuron data from a file and constructs a torch Data object.

We extract the simplified apical trees, and use the apical distances as features.

Parameters

  • path – Path to the neuron file. Supported file types are .h5, .swc, and .pickle.

  • label – The label to be attached to the Data class.

  • nodes_features – List of the nodes features.

  • inverted_apical (boolean) – If True, radial distances are inverted if the mean of the projections is negative.

Returns

A sample of the type Data.

Return type

sample

morphoclass.utils.read_layer_neurons_from_dir(data_dir, layer, labels_dic=None, nodes_features=None, inverted_apical=True)

Read all neurons corresponding to a given layer.

Parameters

  • data_dir – Directory containing subdirectories L[layer]_… with neuron files.

  • layer – The layer for which to load the samples.

  • inverted_apical (boolean) – If True, radial distances are inverted if the mean of the projections is negative.

  • labels_dic (dict) – A mapping from integers to layer names.

  • nodes_features – A list of the nodes features.

Returns

  • samples – A list of samples, each of type Data.

  • labels – A dictionary for assigning integers to layer names.

  • paths – A list of the corresponding paths to the neuron stored in samples.

morphoclass.utils.save_var(var, file_name, exist_ok=False)

Serialize a given variable into a file.

Parameters

  • var (obj) – The variable to be pickled.

  • file_name (str or Path) – The filename where to write the variable.

  • exist_ok (bool) – If set to False existing files won’t be overwritten and an exception will be raised instead.

Raises

FileExistsError – File already exists.

morphoclass.utils.str_time_delta(time_delta, short=True)

Create a nice string representation of a time delta.

The time delta is rounded down to whole seconds.

Parameters

  • time_delta (float) – The time delta. Can be computed as the difference of results of two calls of time.time_perf().

  • short (bool) – Switch between a short and a long string representation.

Returns

time_str – The string representation of the time delta.

Return type

str

morphoclass.utils.suppress_print(suppress_err=False)

Get a context manager for suppressing output to stdout.

morphoclass.utils.tree_to_graph(tree)

Convert a TMD tree to a networkx graph.

Parameters

tree (tmd.Tree.Tree.Tree) –

Returns

The graph representation of the tree.

Return type

networkx.Graph

morphoclass.utils.warn_if_nondeterministic(device)

Create a warning if PyTorch is not in deterministic mode.

Parameters

device (str or torch.device) – The currently used device for which to check the determinism.