Running Reducers

TESS User Reducer

This module porvides functions to calculate uesr weights for the TESS project. Extracts are from Ceasars PluckFieldExtractor.

panoptes_aggregation.running_reducers.tess_user_reducer.tess_user_reducer(data, **kwargs)

Calculate TESS user weights

Parameters:

data (list) – A list with one item containing the extract with the user’s feedback on a gold standard subject
store (keyword, dict) – A dictinary with two keys:
- seed: sum of all previous seed values
- count: sum of all previous gold standard transits seen
relevant_reduction (keyword, list) – A list with one item containing the results of the current subject’s stats reducer. This item is a dictinary with two keys:
- True: number of users who correctly identified the gold standard transits in the subject
- False: number of users who incorrectly identified the gold standard transits in the subject

Returns:

reduction – A dictinary with two keys:

Return type:

dict

This module porvides functions to reduce the column task extracts for the TESS project in running mode. Extracts are from panoptes_aggregation.extractors.shape_extractor.

panoptes_aggregation.running_reducers.tess_reducer_column.tess_reducer_column_rr(data, **kwargs): See panoptes_aggregation.reducers.tess_reducer_column.tess_reducer_column()

This module porvides functions to reduce the gold standard task extracts for the TESS project in running mode.

panoptes_aggregation.running_reducers.tess_gold_standard_reducer.tess_gold_standard_reducer_rr(data, **kwargs): See panoptes_aggregation.reducers.tess_gold_standard_reducer.tess_gold_standard_reducer()

This module provides functions to calculate user weights for the Gravity Spy project. Extracts are from caesar’s PluckFieldExtractor.

panoptes_aggregation.running_reducers.gravity_spy_user_reducer.gravity_spy_user_reducer(data, **kwargs)

Calculate Gravity Spy user weights based on a confusion matrix from gold standard data.

Parameters:

data (list) – A list with one item containing the extract with the user’s choice and the gold standard label.
first_level (str) – A string containing the key for the first level in the level_config object
level_config (dict) – This dictionary holds information about each level in the project. The key must be strings and the values are a dict with up to four keys:
- workflow_id: the workflow ID for the level
- new_categories: the categories added in this level (not included for the final level)
- threshold: the min value of alpha these categories need to trigger a level up (not included for the final level)
- next_level: the key for the next level (not included for the final level). Example:
```
level_config = {
    'level_1': {
        'workflow_id': 1,
        'new_categories': [
            'BLIP',
            'WHISTLE'
        ],
        'threshold': 0.7,
        'next_level': 'level_2'
    },
    'level_2': {
        'workflow_id': 2
    }
}
```
store (keyword, dict) – A dictionary with three keys:
- confusion_matrix: The confusion matrix for the user (stored as nested dict). The first key is the choice given by the user, the second key is the gold standard label.
- column_normalization: The sum of each of the columns (used for normalization). i.e. The total number of time the user has vote for each choice.
- max_level: The maximum workflow level of the user

Returns:

reduction – A dictionary with the following keys:

alpha: A dictionary of values indicating how well the user classifies each category they have seen gold standard images for (diagonal of the normalized confusion matrix).
level_up: Bool indicating if the user should level up (used to trigger effect)
max_workflow_id: The workflow ID for the user’s highest unlocked level
max_level: The maximum workflow level of the user
most_useful_category: The gold standard category the user has the lowest score in (can be used to pick what gold standard category should be shown next to accelerate leveling up)
alpha_length: The number of values in the alpha dict, used to make sure the user has seen every gold standard class of a level before being promoted
normalized_confusion_matrix: The column normalized confusion matrix for the user
_store: The updated store (see above)

Return type:

dict

This module provides functions to calculate subject reductions for the Gravity Spy project. Extracts are from caesar’s PluckFieldExtractor.

panoptes_aggregation.running_reducers.gravity_spy_subject_reducer.gravity_spy_subject_reducer(data, none_key='NONEOFTHEABOVE', **kwargs)

Calculate Gravity Spy category weights for a subject using volunteer’s confusion matrices.

Parameters:

data (list) – A list with one item containing the extract with the user’s choice and the resulting weights from the ML code (stored in the subject metadata)
none_key (string) – The key used for a “none of the above” answer
store (keyword, dict) – A dictionary with two keys:
- number_views: The number of times the subject has been seen (the ML results count for 1 of these)
- category_weights_sum: The running sum for the weights in each category
relevant_reduction (keyword, list) – A list with one item containing the results of the current user’s confusion matrix reducer (see panoptes_aggregation.running_reducers.gravity_spy_user_reduce.gravity_spy_user_reduce())

Returns:

reduction – A dictionary with the following keys:

number_views: Number of times the subject has been seen (the ML results count for 1 of these)
category_weights: A dictionary of values corresponding to the probability the subject belongs to each listed category (all values sum to 1)
max_category_weight: The max value from the category_weights dict, used to retire the subject
_store : The updated store (see above)

Return type:

dict