geowatch.mlops.aggregate_plots module

Plotting logic for Aggregator. These plots illustrate the relationshp between parameters and metrics from various vantage points.

Used by ./aggregate.py

geowatch.mlops.aggregate_plots.build_plotter(agg, rois, plot_config)

Used by Aggregator, which will generally immediately call ParamPlotter.plot_requested().

Returns:

ParamPlotter

geowatch.mlops.aggregate_plots.build_all_param_plots(agg, rois, plot_config)

Main entry point for plotting results from an Aggregator.

geowatch.mlops.aggregate_plots.build_special_columns(agg)

geowatch.mlops.aggregate_plots.preprocess_table_for_seaborn(agg, table)

class geowatch.mlops.aggregate_plots.ParamPlotter(agg, vantage_points=None)

Bases: object

Builds the scatter and box-and-whisker plots over different params. Working on cleaning this up

plot_requested()

Simplified entry point

plot_resources()

Draw tables that summarize the resource usage of the experiments.

plot_overviews()

Draw the overview for each vantage point. Draw tables that summarize the resource usage of the experiments.

plot_params()

plot_vantage_per_region_overview(vantage)

Draw scatter plots and box plots that that distinguish each region with respect to a vantage point.

Parameters:

vantage (Dict) – The primary and secondary metric to look at results with. This must have keys: metric1, metric2 and can optionally contain keys: scale1, scale2, objective1, and objective2.

plot_vantage_macro_overview(vantage)

Draw a scatter plot that gives an overview of the requested macro table wrt to a metric vantage point.

Parameters:

vantage (Dict) – The primary and secondary metric to look at results with. This must have keys: metric1, metric2 and can optionally contain keys: scale1, scale2, objective1, and objective2.

plot_vantage_params(vantage, pman=None, params_of_interest=None)

The main parameter inspection plots.

A vantage point specifies the metrics to visualize and analyze. Makes scatter plot, box plots, and attempts to draw legends useful for communicating results.

The main per-paramter logic is in _plot_single_vantage_param().

Writes two folders in the “macro-plots-*” folder: params, and vantage. Both folders contain the same data, but it is laid out differently.

1. The params folder contains a sub-folder for each parameter, and that folder contains plots for all vantage points.

2. The “vantage” folder contains a sub-folder for each vantage point and all parameters plotted with that vantage point are plotted here. This folder will typically be symlinks into the above “params” folder.

Parameters:

• vantage (Dict) – The primary and secondary metric to look at results with. This must have keys: metric1, metric2 and can optionally contain keys: scale1, scale2, objective1, and objective2.

• pman (ProgressManager | None) – a progress manager used to indicate sub-progress of plotting

geowatch.mlops.aggregate_plots.edit_distance(string1, string2)

Edit distance algorithm. String1 and string2 can be either strings or lists of strings

Parameters:

• string1 (str | List[str])

• string2 (str | List[str])

Requirements:

pip install python-Levenshtein

Returns:

float | List[float] | List[List[float]]

Example

>>> # xdoctest: +REQUIRES(module:Levenshtein)
>>> string1 = 'hello world'
>>> string2 = ['goodbye world', 'rofl', 'hello', 'world', 'lowo']
>>> edit_distance(['hello', 'one'], ['goodbye', 'two'])
>>> edit_distance('hello', ['goodbye', 'two'])
>>> edit_distance(['hello', 'one'], 'goodbye')
>>> edit_distance('hello', 'goodbye')
>>> distmat = edit_distance(string1, string2)
>>> result = ('distmat = %s' % (ub.repr2(distmat),))
>>> print(result)
>>> [7, 9, 6, 6, 7]

geowatch.mlops.aggregate_plots.suggest_did_you_mean(invalid_options, valid_choices)

Parameters:

• missing (List[str]) – the invalid options the user chose

• valid_choices (List[str]) – the valid available options

Example

>>> from geowatch.mlops.aggregate_plots import *  # NOQA
>>> invalid_options = ['fuber', 'yam', 'spam']
>>> valid_choices = ['spam', 'ham', 'foobar']
>>> suggest_did_you_mean(invalid_options, valid_choices)

geowatch.mlops.aggregate_plots.shrink_param_names(param_name, param_values, text_len_thresh=20)

exception geowatch.mlops.aggregate_plots.SkipPlot

Bases: Exception

class geowatch.mlops.aggregate_plots.Vantage2

Bases: dict

The primary and secondary metric to look at results with. This must have keys: metric1, metric2 and can optionally contain keys: scale1, scale2, objective1, and objective2.

property name

property metric1

property metric2

property scale1

property scale2

class geowatch.mlops.aggregate_plots.Vantage(metric1: str, metric2: str, scale1: str = 'linear', scale2: str = 'linear', objective1: str = 'maximize', objective2: str = 'maximize', name: str = None)

Bases: object

The primary and secondary metric to look at results with. This must have keys: metric1, metric2 and can optionally contain keys: scale1, scale2, objective1, and objective2.

Example

>>> from geowatch.mlops.aggregate_plots import *  # NOQA
>>> Vantage('metrics.ppv', 'metrics.tpr')

metric1: str

metric2: str

scale1: str = 'linear'

scale2: str = 'linear'

objective1: str = 'maximize'

objective2: str = 'maximize'

name: str = None