# Copyright 2019 IBM Corporation
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

from sklearn.linear_model import SGDClassifier as SKLModel
import lale.docstrings
import lale.operators

class SGDClassifierImpl():

    def __init__(self, loss='hinge', penalty='l2', alpha=0.0001, l1_ratio=0.15, fit_intercept=True, max_iter=None, tol=None, shuffle=True, verbose=0, epsilon=0.1, n_jobs=None, random_state=None, learning_rate='optimal', eta0=0.0, power_t=0.5, early_stopping=False, validation_fraction=0.1, n_iter_no_change=5, class_weight='balanced', warm_start=False, average=False):
        self._hyperparams = {
            'loss': loss,
            'penalty': penalty,
            'alpha': alpha,
            'l1_ratio': l1_ratio,
            'fit_intercept': fit_intercept,
            'max_iter': max_iter,
            'tol': tol,
            'shuffle': shuffle,
            'verbose': verbose,
            'epsilon': epsilon,
            'n_jobs': n_jobs,
            'random_state': random_state,
            'learning_rate': learning_rate,
            'eta0': eta0,
            'power_t': power_t,
            'early_stopping': early_stopping,
            'validation_fraction': validation_fraction,
            'n_iter_no_change': n_iter_no_change,
            'class_weight': class_weight,
            'warm_start': warm_start,
            'average': average}
        self._wrapped_model = SKLModel(**self._hyperparams)

    def fit(self, X, y=None):
        if (y is not None):
            self._wrapped_model.fit(X, y)
        else:
            self._wrapped_model.fit(X)
        return self

    def predict(self, X):
        return self._wrapped_model.predict(X)

    def predict_proba(self, X):
        return self._wrapped_model.predict_proba(X)

    def decision_function(self, X):
        return self._wrapped_model.decision_function(X)

    def partial_fit(self, X, y=None, classes = None):
      if not hasattr(self, "_wrapped_model"):
        self._wrapped_model = SKLModel(**self._hyperparams)
      self._wrapped_model.partial_fit(X, y, classes = classes)
      return self

_hyperparams_schema = {
    'description': 'inherited docstring for SGDClassifier Linear classifiers (SVM, logistic regression, a.o.) with SGD training.',
    'allOf': [{
        'type': 'object',
        'required': ['loss', 'penalty', 'alpha', 'l1_ratio', 'fit_intercept', 'max_iter', 'tol', 'shuffle', 'verbose', 'epsilon', 'n_jobs', 'random_state', 'learning_rate', 'eta0', 'power_t', 'early_stopping', 'validation_fraction', 'n_iter_no_change', 'class_weight', 'warm_start', 'average'],
        'relevantToOptimizer': ['loss', 'penalty', 'alpha', 'l1_ratio', 'fit_intercept', 'max_iter', 'tol', 'shuffle', 'epsilon', 'learning_rate', 'eta0', 'power_t'],
        'additionalProperties': False,
        'properties': {
            'loss': {
                'enum': ['hinge', 'log', 'modified_huber', 'squared_hinge', 'perceptron', 'squared_loss', 'huber', 'epsilon_insensitive', 'squared_epsilon_insensitive'],
                'default': 'hinge',
                'description': "The loss function to be used. Defaults to 'hinge', which gives a linear SVM."},
            'penalty': {
                'description': "The penalty (aka regularization term) to be used. Defaults to 'l2'",
                'enum': ['elasticnet', 'l1', 'l2'],
                'default': 'l2'},
            'alpha': {
                'type': 'number',
                'minimumForOptimizer': 1e-10,
                'maximumForOptimizer': 1.0,
                'distribution': 'loguniform',
                'default': 0.0001,
                'description': 'Constant that multiplies the regularization term. Defaults to 0.0001'},
            'l1_ratio': {
                'type': 'number',
                'minimumForOptimizer': 1e-9,
                'maximumForOptimizer': 1.0,
                'distribution': 'loguniform',
                'default': 0.15,
                'description': 'The Elastic Net mixing parameter, with 0 <= l1_ratio <= 1.'},
            'fit_intercept': {
                'type': 'boolean',
                'default': True,
                'description': 'Whether the intercept should be estimated or not. If False, the'},
            'max_iter': {
                'type': 'integer',
                'minimumForOptimizer': 10,
                'maximumForOptimizer': 1000,
                'distribution': 'uniform',
                'default': 1000,
                'description': 'The maximum number of passes over the training data (aka epochs).'},
            'tol': {
                'anyOf': [{
                    'type': 'number',
                    'minimumForOptimizer': 1e-08,
                    'maximumForOptimizer': 0.01,
                    'distribution': 'loguniform'}, {
                    'enum': [None]}],
                'default': 0.001,
                'description': 'The stopping criterion.'},
            'shuffle': {
                'type': 'boolean',
                'default': True,
                'description': 'Whether or not the training data should be shuffled after each epoch.'},
            'verbose': {
                'type': 'integer',
                'default': 0,
                'description': 'The verbosity level'},
            'epsilon': {
                'type': 'number',
                'minimumForOptimizer': 1e-08,
                'maximumForOptimizer': 1.35,
                'distribution': 'loguniform',
                'default': 0.1,
                'description': 'Epsilon in the epsilon-insensitive loss functions; only if `loss` is'},
            'n_jobs': {
                'anyOf': [{
                    'type': 'integer'}, {
                    'enum': [None]}],
                'default': None,
                'description': 'The number of CPUs to use to do the OVA (One Versus All, for'},
            'random_state': {
                'anyOf': [
                {   'type': 'integer'},
                {   'laleType': 'numpy.random.RandomState'},
                {   'enum': [None]}],
                'default': None,
                'description': 'The seed of the pseudo random number generator to use when shuffling'},
            'learning_rate': {
                'enum': ['optimal', 'constant', 'invscaling', 'adaptive'],
                'default': 'optimal',
                'description': 'The learning rate schedule:'},
            'eta0': {
                'type': 'number',
                'minimumForOptimizer': 0.01,
                'maximumForOptimizer': 1.0,
                'distribution': 'loguniform',
                'default': 0.0,
                'description': "The initial learning rate for the 'constant', 'invscaling' or"},
            'power_t': {
                'type': 'number',
                'minimumForOptimizer': 0.00001,
                'maximumForOptimizer': 1.0,
                'distribution': 'uniform',
                'default': 0.5,
                'description': 'The exponent for inverse scaling learning rate [default 0.5].'},
            'early_stopping': {
                'type': 'boolean',
                'default': False,
                'description': 'Whether to use early stopping to terminate training when validation'},
            'validation_fraction': {
                'type': 'number',
                'default': 0.1,
                'description': 'The proportion of training data to set aside as validation set for'},
            'n_iter_no_change': {
                'type': 'integer',
                'default': 5,
                'description': 'Number of iterations with no improvement to wait before early stopping.'},
            'class_weight': {
                'anyOf': [{
                    'type': 'object'}, {
                    'enum': ['balanced', None]}],
                'default': None,
                'description': 'Preset for the class_weight fit parameter.'},
            'warm_start': {
                'type': 'boolean',
                'default': False,
                'description': 'When set to True, reuse the solution of the previous call to fit as'},
            'average': {
                'anyOf': [{
                    'type': 'boolean'}, {
                    'type': 'integer'}],
                'default': False,
                'description': 'When set to True, computes the averaged SGD weights and stores the'}
        }}, {
        'description': 'eta0 must be greater than 0 if the learning_rate is not ‘optimal’.',
        'anyOf': [{
            'type': 'object',
            'properties': {
                'learning_rate': {
                    'enum': ['optimal']},
            }}, {
            'type': 'object',
            'properties': {
                'eta0': {
                    'type': 'number',
                    'minimum': 0.0, 
                    'exclusiveMinimum': True },
            }}]}]}

_input_fit_schema = {
    'description': 'Fit linear model with Stochastic Gradient Descent.',
    'required': ['X', 'y'],
    'type': 'object',
    'properties': {
        'X': {
            'type': 'array',
            'items': {
                'type': 'array',
                'items': {
                    'type': 'number'},
            },
            'description': 'Training data'},
        'y': {
            'anyOf': [
                {'type': 'array', 'items': {'type': 'number'}},
                {'type': 'array', 'items': {'type': 'string'}},
                {'type': 'array', 'items': {'type': 'boolean'}}],
            'description': 'Target values'},
        'coef_init': {
            'type': 'array',
            'items': {
                'type': 'array',
                'items': {
                    'type': 'number'},
            },
            'description': 'The initial coefficients to warm-start the optimization.'},
        'intercept_init': {
            'type': 'array',
            'items': {
                'type': 'number'},
            'description': 'The initial intercept to warm-start the optimization.'},
        'sample_weight': {
            'anyOf': [{
                'type': 'array',
                'items': {
                    'type': 'number'},
            }, {
                'enum': [None]}],
            'default': None,
            'description': 'Weights applied to individual samples.'},
    },
}
_input_predict_schema = {
    'description': 'Predict class labels for samples in X.',
    'type': 'object',
    'properties': {
        'X': {
            'type': 'array',
            'items': {
                'type': 'array',
                'items': {
                    'type': 'number'},
            },
            'description': 'Training data'},
    },
}
_output_predict_schema = {
    'description': 'Predicted class label per sample.',
    'anyOf': [
        {'type': 'array', 'items': {'type': 'number'}},
        {'type': 'array', 'items': {'type': 'string'}},
        {'type': 'array', 'items': {'type': 'boolean'}}]}

_input_predict_proba_schema = {
    'description': 'Probability estimates.',
    'type': 'object',
    'properties': {
        'X': {
            'type': 'array',
            'items': {
                'type': 'array',
                'items': {
                    'type': 'number'},
            }},
    },
}
_output_predict_proba_schema = {
    'description': 'Returns the probability of the sample for each class in the model,',
    'type': 'array',
    'items': {
        'type': 'array',
        'items': {
            'type': 'number'},
    }}

_input_decision_function_schema = {
  'type': 'object',
  'required': ['X'],
  'additionalProperties': False,
  'properties': {
    'X': {
      'description': 'Features; the outer array is over samples.',
      'type': 'array',
      'items': {'type': 'array', 'items': {'type': 'number'}}}}}

_output_decision_function_schema = {
    'description': 'Confidence scores for samples for each class in the model.',
    'anyOf': [
    {   'description': 'In the multi-way case, score per (sample, class) combination.',
        'type': 'array',
        'items': {'type': 'array', 'items': {'type': 'number'}}},
    {   'description': 'In the binary case, score for `self._classes[1]`.',
        'type': 'array',
        'items': {'type': 'number'}}]}

_combined_schemas = {
    '$schema': 'http://json-schema.org/draft-04/schema#',
    'description': """`SGD classifier`_ from scikit-learn uses linear classifiers (SVM, logistic regression, a.o.) with stochastic gradient descent training.

.. _`SGD classifier`: https://scikit-learn.org/0.20/modules/generated/sklearn.linear_model.SGDClassifier.html#sklearn-linear-model-sgdclassifier
""",
    'documentation_url': 'https://lale.readthedocs.io/en/latest/modules/lale.lib.sklearn.sgd_classifier.html',
    'import_from': 'sklearn.linear_model',
    'type': 'object',
    'tags': {
        'pre': [],
        'op': ['estimator', 'classifier'],
        'post': []},
    'properties': {
        'hyperparams': _hyperparams_schema,
        'input_fit': _input_fit_schema,
        'input_predict': _input_predict_schema,
        'output_predict': _output_predict_schema,
        'input_predict_proba': _input_predict_proba_schema,
        'output_predict_proba': _output_predict_proba_schema,
        'input_decision_function': _input_decision_function_schema,
        'output_decision_function': _output_decision_function_schema,
}}

lale.docstrings.set_docstrings(SGDClassifierImpl, _combined_schemas)

SGDClassifier = lale.operators.make_operator(SGDClassifierImpl, _combined_schemas)