Usage With State Functionality¶

Installation¶

In [78]:

# Uncomment the following line when running on Google Colab
# !pip install autora
# !pip install autora[synthetic]
# !pip install autora[experimentalist-pprediction-filter]

# Uncomment the following line when running on Google Colab # !pip install autora # !pip install autora[synthetic] # !pip install autora[experimentalist-pprediction-filter]

Imports¶

In [96]:

from autora.state import StandardState, on_state, Delta, estimator_on_state
from autora.variable import Variable, VariableCollection
from autora.experimentalist.random import random_pool
from autora.experimentalist.prediction_filter import prediction_filter
from autora.experiment_runner.synthetic.psychology.luce_choice_ratio import luce_choice_ratio
from sklearn.linear_model import LinearRegression
from sklearn.linear_model import LogisticRegression

from autora.state import StandardState, on_state, Delta, estimator_on_state from autora.variable import Variable, VariableCollection from autora.experimentalist.random import random_pool from autora.experimentalist.prediction_filter import prediction_filter from autora.experiment_runner.synthetic.psychology.luce_choice_ratio import luce_choice_ratio from sklearn.linear_model import LinearRegression from sklearn.linear_model import LogisticRegression

Setup: Experiment¶

In [80]:

experiment = luce_choice_ratio()

experiment = luce_choice_ratio()

Setup: State¶

In [81]:

state = StandardState(experiment.variables)
state.variables

state = StandardState(experiment.variables) state.variables

Out[81]:

VariableCollection(independent_variables=[IV(name='similarity_category_A1', value_range=(0.1, 10), allowed_values=array([ 0.1       ,  1.51428571,  2.92857143,  4.34285714,  5.75714286,
        7.17142857,  8.58571429, 10.        ]), units='similarity', type=<ValueType.REAL: 'real'>, variable_label='Similarity with Category A1', rescale=1, is_covariate=False), IV(name='similarity_category_A2', value_range=(0.1, 10), allowed_values=array([ 0.1       ,  1.51428571,  2.92857143,  4.34285714,  5.75714286,
        7.17142857,  8.58571429, 10.        ]), units='similarity', type=<ValueType.REAL: 'real'>, variable_label='Similarity with Category A2', rescale=1, is_covariate=False), IV(name='similarity_category_B1', value_range=(0.1, 10), allowed_values=array([ 0.1       ,  1.51428571,  2.92857143,  4.34285714,  5.75714286,
        7.17142857,  8.58571429, 10.        ]), units='similarity', type=<ValueType.REAL: 'real'>, variable_label='Similarity with Category B1', rescale=1, is_covariate=False), IV(name='similarity_category_B2', value_range=(0.1, 10), allowed_values=array([ 0.1       ,  1.51428571,  2.92857143,  4.34285714,  5.75714286,
        7.17142857,  8.58571429, 10.        ]), units='similarity', type=<ValueType.REAL: 'real'>, variable_label='Similarity with Category B2', rescale=1, is_covariate=False)], dependent_variables=[DV(name='choose_A1', value_range=(0, 1), allowed_values=None, units='probability', type=<ValueType.PROBABILITY: 'probability'>, variable_label='Probability of Choosing A1', rescale=1, is_covariate=False)], covariates=[])

Experimentalist: Random Pool on State¶

In [82]:

@on_state
def pool_on_state(variables, num_samples, random_state=42):
    return Delta(conditions=random_pool(
        variables=variables, 
        num_samples=num_samples, 
        random_state=random_state)
    )
state = pool_on_state(state, num_samples=100)
state.conditions

@on_state def pool_on_state(variables, num_samples, random_state=42): return Delta(conditions=random_pool( variables=variables, num_samples=num_samples, random_state=random_state) ) state = pool_on_state(state, num_samples=100) state.conditions

Out[82]:

	similarity_category_A1	similarity_category_A2	similarity_category_B1	similarity_category_B2
0	0.100000	8.585714	2.928571	2.928571
1	8.585714	1.514286	10.000000	8.585714
2	7.171429	8.585714	4.342857	0.100000
3	4.342857	0.100000	7.171429	10.000000
4	4.342857	8.585714	4.342857	5.757143
...	...	...	...	...
95	2.928571	0.100000	5.757143	8.585714
96	1.514286	2.928571	8.585714	0.100000
97	7.171429	0.100000	4.342857	1.514286
98	7.171429	2.928571	2.928571	2.928571
99	1.514286	10.000000	0.100000	2.928571

100 rows × 4 columns

Experiment Runner: Run Experiment¶

In [83]:

@on_state
def run_on_state(conditions, random_state=42):
    return Delta(experiment_data=experiment.run(conditions, random_state=random_state))
state = run_on_state(state)
state.experiment_data

@on_state def run_on_state(conditions, random_state=42): return Delta(experiment_data=experiment.run(conditions, random_state=random_state)) state = run_on_state(state) state.experiment_data

Out[83]:

	similarity_category_A1	similarity_category_A2	similarity_category_B1	similarity_category_B2	choose_A1
0	0.100000	8.585714	2.928571	2.928571	0.010227
1	8.585714	1.514286	10.000000	8.585714	0.581342
2	7.171429	8.585714	4.342857	0.100000	0.425710
3	4.342857	0.100000	7.171429	10.000000	0.498484
4	4.342857	8.585714	4.342857	5.757143	0.279448
...	...	...	...	...	...
95	2.928571	0.100000	5.757143	8.585714	0.440030
96	1.514286	2.928571	8.585714	0.100000	0.226442
97	7.171429	0.100000	4.342857	1.514286	0.819505
98	7.171429	2.928571	2.928571	2.928571	0.620568
99	1.514286	10.000000	0.100000	2.928571	0.122477

100 rows × 5 columns

Theorist: Linear Regression

In [101]:

@on_state
def theorist_on_state(X, y):
    return Delta(models=[LinearRegression().fit(X, y)])
state = theorist_on_state(state)
state.models[-1]

@on_state def theorist_on_state(X, y): return Delta(models=[LinearRegression().fit(X, y)]) state = theorist_on_state(state) state.models[-1]

Out[101]:

LinearRegression()

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

Filter: Prediction Filter¶

Let's look at predictions first:

In [102]:

conditions = random_pool(variables=experiment.variables, num_samples=4, random_state=42)
state.models[-1].predict(conditions)

conditions = random_pool(variables=experiment.variables, num_samples=4, random_state=42) state.models[-1].predict(conditions)

Out[102]:

array([[0.11630747],
       [0.4675618 ],
       [0.63628483],
       [0.21396789]])

Here, we want predictions that are either smaller then .2 or larger then .8

In [103]:

filter_function = lambda x : x < .2 or x > .8

filter_function = lambda x : x < .2 or x > .8

In [104]:

@on_state
def filter_on_state(conditions, models):
    filtered = prediction_filter(conditions, models[-1], filter_function)
    return Delta(conditions=filtered)
state = pool_on_state(state, num_samples=10, random_state=42)
state = filter_on_state(state)
state.conditions

@on_state def filter_on_state(conditions, models): filtered = prediction_filter(conditions, models[-1], filter_function) return Delta(conditions=filtered) state = pool_on_state(state, num_samples=10, random_state=42) state = filter_on_state(state) state.conditions

Out[104]:

	similarity_category_A1	similarity_category_A2	similarity_category_B1	similarity_category_B2
0	0.100000	5.757143	5.757143	4.342857
1	4.342857	8.585714	10.000000	5.757143
2	0.100000	5.757143	4.342857	8.585714
3	1.514286	8.585714	5.757143	2.928571
4	0.100000	4.342857	4.342857	7.171429

Let's look at the prediction to validate our filter

In [105]:

state.models[-1].predict(state.conditions)

state.models[-1].predict(state.conditions)

Out[105]:

array([[0.07178846],
       [0.1923867 ],
       [0.04447933],
       [0.07532331],
       [0.10101559]])

Let's run the experiment on these conditions and look at

In [106]:

state = run_on_state(state)
state.y[-5:]

state = run_on_state(state) state.y[-5:]

Out[106]:

	choose_A1
130	0.012385
131	0.256693
132	0.012034
133	0.124357
134	0.010328

In [106]:

In [ ]: