InteractivePlots

class fedartml.InteractivePlots(labels, random_state=None, colors=None, distance='jensen-shannon', **plot_kwargs)

Generate simulated interactive plots (with sliders) from the labels provided in a federated learning paradigm to exemplify identically and non-identically distributed labels across the local nodes (clients).

Parameters

labelsarray-like

The target values (class labels in classification).

random_stateint

Controls the shuffling applied to the generation of pseudorandom numbers. Pass an int for reproducible output across multiple function calls.

colorslist

Colors list used to plot. Must have a length of 7 positions.

distancestr

Distance to use for measuring heterogeneity (non-IID-ness) of the label’s distribution among clients. Possible choices: “jensen-shannon”, “hellinger”, “earth_movers”.

**plot_kwargsdict

Keyword arguments used for customizing plots (inherited from matplotlib.pyplot).

bar_divided_distr_dirichlet(Alpha, Local_Nodes)

Create an interactive bar plot (with sliders) divided per each local node (client).

Parameters

Alphaslider

Concentration parameter of the Dirichlet distribution.

Local_Nodesslider

Number of local nodes (clients) used in the federated learning paradigm.

Returns

The return keyword is empty. The function shows the plot as output.

bar_divided_distr_percent_noniid(Pctg_NonIID, Local_Nodes)

Create an interactive bar plot (with sliders) divided per each local node (client).

Parameters

Pctg_NonIIDslider

Percentage (between o and 100) desired of non-IID-ness for the federated data.

Local_Nodesslider

Number of local nodes (clients) used in the federated learning paradigm.

Returns

The return keyword is empty. The function shows the plot as output.

scatter_distr_dirichlet(Alpha, Local_Nodes)

Create an interactive scatter plot (with sliders) per each local node (client) and label’s classes.

Parameters

Alphaslider

Concentration parameter of the Dirichlet distribution.

Local_Nodesslider

Number of local nodes (clients) used in the federated learning paradigm.

Returns

The return keyword is empty. The function shows the plot as output.

scatter_distr_percent_noniid(Pctg_NonIID, Local_Nodes)

Create an interactive scatter plot (with sliders) per each local node (client) and label’s classes.

Parameters

Pctg_NonIIDslider

Percentage (between o and 100) desired of non-IID-ness for the federated data.

Local_Nodesslider

Number of local nodes (clients) used in the federated learning paradigm.

Returns

The return keyword is empty. The function shows the plot as output.

show_bar_divided_distr_dirichlet(**slider_kwargs)

Show an interactive bar plot (with sliders) divided per each local node (client).

Parameters

**slider_kwargs: dict

Keyword arguments used for customizing sliders (inherited from ipywidgets.interact).

Returns

The return keyword is empty. The function shows the sliders for Alpha and number of local nodes (clients).

References

[1]

Tao Lin∗, Lingjing Kong∗, Sebastian U. Stich, Martin Jaggi. (2020). Ensemble Distillation for Robust Model Fusion in Federated Learning https://proceedings.neurips.cc/paper/2020/file/18df51b97ccd68128e994804f3eccc87-Supplemental.pdf

Examples

>>> from fedartml import InteractivePlots
>>> from keras.datasets import mnist
>>> (train_X, train_y), (test_X, test_y) = mnist.load_data()
>>> my_plot = InteractivePlots(labels = train_y)
>>> my_plot.show_bar_divided_distr_dirichlet()

show_bar_divided_distr_percent_noniid(**slider_kwargs)

Show an interactive bar plot (with sliders) divided per each local node (client).

Parameters

**slider_kwargs: dict

Keyword arguments used for customizing sliders (inherited from ipywidgets.interact).

Returns

The return keyword is empty. The function shows the sliders for Pctg_NonIID and number of local nodes (clients).

References

[1]

Hsieh, K., Phanishayee, A., Mutlu, O., & Gibbons, P. (2020, November). The non-iid data quagmire of decentralized machine learning. In International Conference on Machine Learning (pp. 4387-4398). PMLR. https://proceedings.mlr.press/v119/hsieh20a/hsieh20a.pdf

Examples

>>> from fedartml import InteractivePlots
>>> from keras.datasets import mnist
>>> (train_X, train_y), (test_X, test_y) = mnist.load_data()
>>> my_plot = InteractivePlots(labels = train_y)
>>> my_plot.show_bar_divided_distr_percent_noniid()

show_scatter_distr_dirichlet(**slider_kwargs)

Show an interactive scatter plot (with sliders) per each local node (client) and label’s classes.

Parameters

**slider_kwargs: dict

Keyword arguments used for customizing sliders (inherited from ipywidgets.interact).

Returns

The return keyword is empty. The function shows the sliders for Alpha and number of local nodes (clients).

References

[1]

Tao Lin∗, Lingjing Kong∗, Sebastian U. Stich, Martin Jaggi. (2020). Ensemble Distillation for Robust Model Fusion in Federated Learning https://proceedings.neurips.cc/paper/2020/file/18df51b97ccd68128e994804f3eccc87-Supplemental.pdf

Examples

>>> from fedartml import InteractivePlots
>>> from keras.datasets import mnist
>>> (train_X, train_y), (test_X, test_y) = mnist.load_data()
>>> my_plot = InteractivePlots(labels = train_y)
>>> my_plot.show_scatter_distr_dirichlet()

show_scatter_distr_percent_noniid(**slider_kwargs)

Show an interactive scatter plot (with sliders) per each local node (client) and label’s classes.

Parameters

**slider_kwargs: dict

Keyword arguments used for customizing sliders (inherited from ipywidgets.interact).

Returns

The return keyword is empty. The function shows the sliders for Pctg_NonIID and number of local nodes (clients).

References

[1]

Hsieh, K., Phanishayee, A., Mutlu, O., & Gibbons, P. (2020, November). The non-iid data quagmire of decentralized machine learning. In International Conference on Machine Learning (pp. 4387-4398). PMLR. https://proceedings.mlr.press/v119/hsieh20a/hsieh20a.pdf

Examples

>>> from fedartml import InteractivePlots
>>> from keras.datasets import mnist
>>> (train_X, train_y), (test_X, test_y) = mnist.load_data()
>>> my_plot = InteractivePlots(labels = train_y)
>>> my_plot.show_scatter_distr_percent_noniid()

show_stacked_distr_dirichlet(**slider_kwargs)

Show an interactive stacked bar plot (with sliders) per each local node (client) and label’s classes.

Parameters

**slider_kwargs: dict

Keyword arguments used for customizing sliders (inherited from ipywidgets.interact).

Returns

The return keyword is empty. The function shows the sliders for Alpha and number of local nodes (clients).

References

[1]

Tao Lin∗, Lingjing Kong∗, Sebastian U. Stich, Martin Jaggi. (2020). Ensemble Distillation for Robust Model Fusion in Federated Learning https://proceedings.neurips.cc/paper/2020/file/18df51b97ccd68128e994804f3eccc87-Supplemental.pdf

Examples

>>> from fedartml import InteractivePlots
>>> from keras.datasets import mnist
>>> (train_X, train_y), (test_X, test_y) = mnist.load_data()
>>> my_plot = InteractivePlots(labels = train_y)
>>> my_plot.show_stacked_distr_dirichlet()

show_stacked_distr_percent_noniid(**slider_kwargs)

Show an interactive stacked bar plot (with sliders) per each local node (client) and label’s classes.

Parameters

**slider_kwargs: dict

Keyword arguments used for customizing sliders (inherited from ipywidgets.interact).

Returns

The return keyword is empty. The function shows the sliders for Pctg_noniid and number of local nodes (clients).

References

[1]

Hsieh, K., Phanishayee, A., Mutlu, O., & Gibbons, P. (2020, November). The non-iid data quagmire of decentralized machine learning. In International Conference on Machine Learning (pp. 4387-4398). PMLR. https://proceedings.mlr.press/v119/hsieh20a/hsieh20a.pdf

Examples

>>> from fedartml import InteractivePlots
>>> from keras.datasets import mnist
>>> (train_X, train_y), (test_X, test_y) = mnist.load_data()
>>> my_plot = InteractivePlots(labels = train_y)
>>> my_plot.show_stacked_distr_percent_noniid()

stacked_distr_dirichlet(Alpha, Local_Nodes)

Create an interactive stacked bar plot (with sliders) per each local node (client) and label’s classes.

Parameters

Alphaslider

Concentration parameter of the Dirichlet distribution.

Local_Nodesslider

Number of local nodes (clients) used in the federated learning paradigm.

Returns

The return keyword is empty. The function shows the plot as output.

stacked_distr_percent_noniid(Pctg_NonIID, Local_Nodes)

Create an interactive stacked bar plot (with sliders) per each local node (client) and label’s classes.

Parameters

Pctg_NonIIDslider

Percentage (between o and 100) desired of non-IID-ness for the federated data.

Local_Nodesslider

Number of local nodes (clients) used in the federated learning paradigm.

Returns

The return keyword is empty. The function shows the plot as output.