tnmf.utils.demo

Provides certain utilities for the streamlit demo.

Module Contents

Classes

SignalTool

An abstract base class that serves as a factory for creating specialized objects that facilitate the handling of

SignalTool1D

A utility class to handle 1-D multi-channel signals (time series data).

SignalTool2D

A utility class to handle 2-D multi-channel signals (image data).

Functions

explanation(text: str, verbose: bool)

st_define_nmf_params(default_params: dict, have_ground_truth: bool = True, verbose: bool = True) → dict

Defines all necessary NMF parameters via streamlit widgets.

fit_nmf_model(V, nmf_params, fit_params, progress_bar)

Attributes

HELP_CHANNEL

tnmf.utils.demo.HELP_CHANNEL = Multiline-String
Show Value
1
2
3
The **number of channels** each input signal comprises. In contrast to the remaining signal dimensions, which are
    treated shift-invariant (meaning that atom placement is flexible), channels represent the inflexible part of the
    factorization in the sense that each atom always covers all channels.
tnmf.utils.demo.explanation(text: str, verbose: bool)
tnmf.utils.demo.st_define_nmf_params(default_params: dict, have_ground_truth: bool = True, verbose: bool = True) → dict

Defines all necessary NMF parameters via streamlit widgets.

Parameters

  • default_params (dict) – Contains the default parameters that are used if the created streamlit checkbox is True.

  • have_ground_truth (bool) – If True, the parameters in default_params are considered being ground truth values and the respective explanatory texts are modified accordingly.

  • verbose (bool) – If True, show detailed information.

Returns

nmf_params – A dictionary containing the selected NMF parameters.

Return type

dict

class tnmf.utils.demo.SignalTool

Bases: abc.ABC

An abstract base class that serves as a factory for creating specialized objects that facilitate the handling of different signal types.

classmethod st_generate_input(cls, verbose: bool = True) → Tuple[numpy.ndarray, dict]

Defines all signal parameters via streamlit widgets and returns a generated input matrix V for the NMF together with a dictionary containing details of the used NMF atoms.

Parameters

verbose (bool) – If True, show detailed information.

Returns

  • V (np.ndarray) – The generated input for the NMF.

  • nmf_params (dict) – Ground truth NMF atom parameters that were used for the signal generation.

classmethod st_compare_signals(cls, V: numpy.ndarray, R: numpy.ndarray, verbose: bool = True)

Compares a given input matrix with its NMF reconstruction in streamlit.

Parameters

  • V (np.ndarray) – The input that was factorized via NMF.

  • R (np.ndarray) – The NMF reconstruction of the input.

  • verbose (bool) – If True, show detailed information.

classmethod st_compare_individual_signals(cls, V: numpy.ndarray, R: numpy.ndarray, verbose: bool = True)

Selects a particular signal and its reconstruction from the given input via a streamlit widget and compares them.

Parameters

  • V (np.ndarray) – The input that was factorized via NMF.

  • R (np.ndarray) – The NMF reconstruction of the input.

  • verbose (bool) – If True, show detailed information.

classmethod st_plot_partial_reconstructions(cls, V: numpy.ndarray, nmf: tnmf.TransformInvariantNMF.TransformInvariantNMF, verbose: bool = True)

Visualizes the partial reconstructions of the given input by the different NMF atoms.

Parameters

  • V (np.ndarray) – The input that was factorized via NMF.

  • nmf (TransformInvariantNMF) – The trained NMF object.

  • verbose (bool) – If True, show detailed information.

classmethod plot_signals(cls, signals: List[numpy.ndarray], signal_opts: Optional[Iterable[dict]] = None, opts: Optional[dict] = None)

Wrapper for _plot_signals to fill the default arguments.

abstract classmethod st_define_signal_params(cls, verbose: bool = True) → dict

Defines all signal parameters via streamlit widgets and returns them in a dictionary.

Parameters

verbose (bool) – If True, show detailed information.

abstract classmethod generate_signal(cls, signal_params: dict) → Tuple[numpy.ndarray, numpy.ndarray]

Creates a single signal using the specified signal parameters. Returns the signal and the used NMF atoms.

class tnmf.utils.demo.SignalTool1D

Bases: SignalTool

A utility class to handle 1-D multi-channel signals (time series data).

classmethod st_define_signal_params(cls, verbose: bool = True) → dict

Defines all signal parameters via streamlit widgets and returns them in a dictionary.

Parameters

verbose (bool) – If True, show detailed information.

classmethod generate_signal(cls, signal_params: dict) → Tuple[numpy.ndarray, numpy.ndarray]

Creates a single signal using the specified signal parameters. Returns the signal and the used NMF atoms.

class tnmf.utils.demo.SignalTool2D

Bases: SignalTool

A utility class to handle 2-D multi-channel signals (image data).

classmethod st_define_signal_params(cls, verbose: bool = True) → dict

Defines all signal parameters via streamlit widgets and returns them in a dictionary.

Parameters

verbose (bool) – If True, show detailed information.

classmethod generate_signal(cls, signal_params: dict) → Tuple[numpy.ndarray, numpy.ndarray]

Creates a single signal using the specified signal parameters. Returns the signal and the used NMF atoms.

tnmf.utils.demo.fit_nmf_model(V, nmf_params, fit_params, progress_bar)