experiment.hooks package

Submodules

experiment.hooks.hook_skeleton module

class experiment.hooks.hook_skeleton.CompositeHook(hook_list)

Bases: object

A hook that calls other hooks.

__init__(hook_list)

Constructor for CompositeHook.

Parameters

hook_list (List[GSSLHook]) – The list of hooks forming this CompositeHook.

class experiment.hooks.hook_skeleton.GSSLHook

Bases: object

Skeleton for GSSL hooks (for plotting, for example).

get_step_rate()

Specifies how many steps it takes to call the “step” procedure on this hook.

experiment.hooks.ldst_filterstats_hook module

Created on 3 de abr de 2019

@author: klaus

class experiment.hooks.ldst_filterstats_hook.LDSTFilterStatsHook(experiment, step_size=5)

Bases: experiment.hooks.hook_skeleton.GSSLHook

Hook that saves stats of the filter at each step.

__init__(experiment, step_size=5)

Constructor for LDSTFilterStatsHook.

Parameters

  • experiment (experiment.experiments.Experiment) – The Experiment object. The experiment’s output_dict will be modified.

  • step_size (int) – Determines after how many iterations to save the filter stats.

best_f1 = -1.0

experiment.hooks.plot_hooks module

Created on 2 de abr de 2019

@author: klaus

class experiment.hooks.plot_hooks.plotHook(filename_path, title, experiment, when='begin', only_labeled=True, force_Y_callback=None, force_lb_callback=None, plot_mode='discrete', palette=None)

Bases: experiment.hooks.hook_skeleton.GSSLHook

Hook that plots labels. Uses a callback to an Experiment object to get X and W, if available.

__init__(filename_path, title, experiment, when='begin', only_labeled=True, force_Y_callback=None, force_lb_callback=None, plot_mode='discrete', palette=None)

Constructor for the plotHook.

Parameters

  • filename_path (str) – Path to the output file (including filename and extension).

  • title (str) – The title to be displayed on the plot.

  • experiment (experiment.experiments.Experiment) – The Experiment object, used for callbacks.

  • when (str) – Either ‘begin’ or ‘end’. If it is ‘begin’, plotting occurs when ‘_begin’ method is called. Otherwise, it occurs when ‘_end’ is called. Default is 'begin'.

  • only_labeled (bool) – If True, future calls must have ‘labeledIndexes’ kwarg, and data marked as unlabeled will have a different color. Default is True.

  • force_Y_callback (str) – Optional. Should be an attribute from experiment.experiments.Experiment that will override the Y variable (which determines the color of the plot).

  • plot_mode (str) – Either 'discrete'` or 'continuous'. The plot type. Default is 'discrete'.

  • palette (str) – Optional. Overrides the palette setting for the plot.

plot(**kwargs)
class experiment.hooks.plot_hooks.plotIterGTAMHook(video_path, title, experiment, step_size=10, mode='Y', palette=None, keep_images=False, create_video=True)

Bases: experiment.hooks.plot_hooks.plotIterHook

Hook that is specific to the GTAM algorithm. There are 3 modes: 1. 'Y': Plots the updated initial belief matrix for each iteration. 2. 'Q': Plots the argmin of each row of the gradient. The lower the values, the more the cost function is decreased

with the labeling of that instance.

  1. 'F': Plots the updated classification for each iteration.

__init__(video_path, title, experiment, step_size=10, mode='Y', palette=None, keep_images=False, create_video=True)

Constructor for the plotIterGTAMHook.

Parameters

  • video_path (str) – Path to the output video (including filename and ‘.mp4’ extension).

  • title (str) – The title to be displayed on the plot.

  • experiment (experiment.experiments.Experiment) – The Experiment object, used for callbacks.

  • step_size (int) – Determines after how many steps should a plotting operation occur.

  • mode (str) – The mode of this plotIterGTAMHook. Must be in ['Y','Q','F']. Default is Y.

  • palette (str) – Optional. Overrides the palette setting for the plot.

  • keep_images (bool) – Whether to keep the images. Default is False.

  • create_video (bool) – Whether to create a video showcasing a slideshow of the images. Default is True.

class experiment.hooks.plot_hooks.plotIterHook(video_path, title, experiment, only_labeled=True, step_size=5, force_Y_callback=None, force_lb_callback=None, temp_subfolder_name='iter', plot_mode='discrete', palette=None, keep_images=False, create_video=True, slowdown_factor=1.0, plot_id=0)

Bases: experiment.hooks.plot_hooks.plotHook

Hook that plots the labels iteratively. Uses a callback to an Experiment object to get X and W, if available.

__init__(video_path, title, experiment, only_labeled=True, step_size=5, force_Y_callback=None, force_lb_callback=None, temp_subfolder_name='iter', plot_mode='discrete', palette=None, keep_images=False, create_video=True, slowdown_factor=1.0, plot_id=0)

Constructor for the plotHook.

Parameters

  • video_path (str) – Path to the output video (including filename and ‘.mp4’ extension).

  • title (str) – The title to be displayed on the plot.

  • experiment (experiment.experiments.Experiment) – The Experiment object, used for callbacks.

  • only_labeled (bool) – If True, future calls must have ‘labeledIndexes’ kwarg, and data marked as unlabeled will have a different color. Default is True.

  • step_size (int) – Determines after how many steps should a plotting operation occur.

  • force_Y_callback (str) – Optional. Should be an attribute from experiment.experiments.Experiment that will override the Y variable (which determines the color of the plot).

  • temp_subfolder_name (str) – Determines the name of the temporary subfolder to store the plot slideshow images. Default is 'iter'.

  • plot_mode (str) – Either 'discrete'` or 'continuous'. The plot type. Default is 'discrete'.

  • palette (str) – Optional. Overrides the palette setting for the plot.

createVideo()
rmFolders()

experiment.hooks.time_hook module

Created on 2 de abr de 2019

@author: klaus

class experiment.hooks.time_hook.timeHook(experiment, timer_name)

Bases: experiment.hooks.hook_skeleton.GSSLHook

Hook that times some part of the experiment execution.

__init__(experiment, timer_name)

Constructor for the timeHook.

Parameters

  • experiment (experiment.experiments.Experiment) – The Experiment object. The experiment’s output_dict will be modified.

  • timer_name (str) – The name of the timer. This name will be a key added to experiment.output_dict. Initially the corresponding value will be np.nan, but at the end it will be updated to be the time taken between _begin and _end calls.

Module contents