MLPRegressor

class hana_ml.algorithms.pal.neural_network.MLPRegressor(activation=None, activation_options=None, output_activation=None, output_activation_options=None, hidden_layer_size=None, hidden_layer_size_options=None, max_iter=None, training_style='stochastic', learning_rate=None, momentum=None, batch_size=None, normalization=None, weight_init=None, categorical_variable=None, resampling_method=None, evaluation_metric=None, fold_num=None, repeat_times=None, search_strategy=None, random_search_times=None, random_state=None, timeout=None, progress_indicator_id=None, param_values=None, param_range=None, thread_ratio=None, reduction_rate=None, aggressive_elimination=None)

Multi-layer perceptron (MLP) Regressor.

Parameters:

activationstr

Specifies the activation function for the hidden layer.

Valid activation functions include:

'tanh',
'linear',
'sigmoid_asymmetric',
'sigmoid_symmetric',
'gaussian_asymmetric',
'gaussian_symmetric',
'elliot_asymmetric',
'elliot_symmetric',
'sin_asymmetric',
'sin_symmetric',
'cos_asymmetric',
'cos_symmetric',
'relu'

Should not be specified only if activation_options is provided.

activation_optionslist of str, optional

A list of activation functions for parameter selection.

See activation for the full set of valid activation functions.

output_activationstr

Specifies the activation function for the output layer.

Valid choices of activation function same as those in activation.

Should not be specified only if output_activation_options is provided.

output_activation_optionslist of str, conditionally mandatory

A list of activation functions for the output layer for parameter selection.

See activation for the full set of activation functions for output layer.

hidden_layer_sizelist of int or tuple of int

Sizes of all hidden layers.

Should not be specified only if hidden_layer_size_options is provided.

hidden_layer_size_optionslist of tuples, optional

A list of optional sizes of all hidden layers for parameter selection.

max_iterint, optional

Maximum number of iterations.

Defaults to 100.

training_style{'batch', 'stochastic'}, optional

Specifies the training style.

Defaults to 'stochastic'.

learning_ratefloat, optional

Specifies the learning rate.

Mandatory and valid only when training_style is 'stochastic'.

momentumfloat, optional

Specifies the momentum for gradient descent update.

Mandatory and valid only when training_style is 'stochastic'.

batch_sizeint, optional

Specifies the size of mini batch.

Valid only when training_style is 'stochastic'.

Defaults to 1.

normalization{'no', 'z-transform', 'scalar'}, optional

Defaults to 'no'.

weight_init{'all-zeros', 'normal', 'uniform', 'variance-scale-normal', 'variance-scale-uniform'}, optional

Specifies the weight initial value.

Defaults to 'all-zeros'.

categorical_variablestr or a list of str, optional

Specifies column name(s) in the data table used as category variable.

Valid only when column is of INTEGER type.

thread_ratiofloat, optional

Controls the proportion of available threads to use for training.

The value range is from 0 to 1, where 0 indicates a single thread, and 1 indicates up to all available threads.

Values between 0 and 1 will use that percentage of available threads.

Values outside this range tell PAL to heuristically determine the number of threads to use.

Defaults to 0.

resampling_methodstr, optional

Specifies the resampling method for model evaluation or parameter selection. Valid options are listed as follows: 'cv', 'bootstrap', 'cv_sha', 'bootstrap_sha', 'cv_hyperband', 'bootstrap_hyperband'.

If not specified, neither model evaluation or parameter selection shall be triggered.

Note

Resampling methods with suffix 'sha' or 'hyperband' are for parameter selection only, not for model evaluation.

evaluation_metric{'rmse'}, optional

Specifies the evaluation metric for model evaluation or parameter selection. Must be specified together with resampling_method to activate model evaluation or parameter selection.

No default value.

fold_numint, optional

Specifies the fold number for the cross-validation.

Mandatory and valid only when resampling_method is specified as one of the following: 'cv', 'cv_sha', 'cv_hyperband'.

repeat_timesint, optional

Specifies the number of repeat times for resampling.

Defaults to 1.

search_strategy{'grid', 'random'}, optional

Specifies the method for parameter selection.

if resampling_method is specified as 'cv_sha' or 'bootstrap_sha', then this parameter is mandatory.
if resampling_method is specified as 'cv_hyperband' or 'bootstrap_hyperband', then this parameter defaults to 'random' and cannot be changed.
otherwise this parameter has no default value, and parameter selection will not be activated if it is not specified.

random_searhc_timesint, optional

Specifies the number of times to randomly select candidate parameters.

Mandatory and valid only when search_strategy is set to 'random'.

random_stateint, optional

Specifies the seed for random generation.

When 0 is specified, system time is used.

Defaults to 0.

timeoutint, optional

Specifies maximum running time for model evaluation/parameter selection, in seconds.

No timeout when 0 is specified.

Defaults to 0.

progress_idstr, optional

Sets an ID of progress indicator for model evaluation/parameter selection.

If not provided, no progress indicator is activated.

param_valuesdict or list of tuples, optional

Specifies the values of following parameters for model parameter selection:

learning_rate, momentum, batch_size.

If input is list of tuples, then each tuple must contain exactly two elements:

1st element is the parameter name(str type),

2nd element is a list of valid values for that parameter.

Otherwise, if input is dict, then for each element, the key must be a parameter name, while value be a list of valid values for that parameter.

A simple example for illustration:

[('learning_rate', [0.1, 0.2, 0.5]), ('momentum', [0.2, 0.6])],

or

dict(learning_rate=[0.1, 0.2, 0.5], momentum=[0.2, 0.6]).

Valid only when resampling_method and search_strategy are both specified, and training_style is 'stochastic'.

param_rangedict or list of tuple, optional

Sets the range of the following parameters for model parameter selection:

learning_rate, momentum, batch_size.

If input is a list of tuples, the each tuple should contain exactly two elements:

1st element is the parameter name(str type),

2nd element is a list that specifies the range of that parameter as follows: first value is the start value, second value is the step, and third value is the end value. The step value can be omitted, and will be ignored, if search_strategy is set to 'random'.

Otherwise, if input is a dict, then for each element the key should be parameter name, while value specifies the range of that parameter.

Valid only when resampling_method and search_strategy are both specified, and training_style is 'stochastic'.

reduction_ratefloat, optional

Specifies reduction rate in SHA or Hyperband method.

For each round, the available parameter candidate size will be divided by value of this parameter. Thus valid value for this parameter must be greater than 1.0

Valid only when resampling_method takes one of the following values: 'cv_sha', 'bootstrap_sha', 'cv_hyperband', 'bootstrap_hyperband'.

Defaults to 3.0.

aggressive_eliminationbool, optional

Specifies whether to apply aggressive elimination while using SHA method.

Aggressive elimination happens when the data size and parameters size to be searched does not match and there are still bunch of parameters to be searched while data size reaches its upper limits. If aggressive elimination is applied, lower bound of limit of data size will be used multiple times first to reduce number of parameters.

Valid only when resampling_method is 'cv_sha' or 'bootstrap_sha'.

Defaults to False.

Examples

Training data df:

>>> df.collect()
   V000  V001 V002  V003  T001  T002  T003
0     1  1.71   AC     0  12.7   2.8  3.06
1    10  1.78   CA     5  12.1   8.0  2.65
...
8    12  2.13   AC     4  13.2   1.9  1.34
9    18  1.87   AC     6  25.5   3.6  2.14

Create a MLPRegressor instance and call fit():

>>> mlpr = MLPRegressor(hidden_layer_size=(10,5),
...                     activation='sin_asymmetric',
...                     output_activation='sin_asymmetric',
...                     learning_rate=0.001, momentum=0.00001,
...                     training_style='batch',
...                     max_iter=10000, normalization='z-transform',
...                     weight_init='normal', thread_ratio=0.3)
>>> mlpr.fit(data=df, label=['T001', 'T002', 'T003'])

Training result may look different from the following results due to model randomness.

>>> mlpr.model_.collect()
   ROW_INDEX                                      MODEL_CONTENT
0          1  {"CurrentVersion":"1.0","DataDictionary":[{"da...
1          2  3782583596893},{"from":10,"weight":-0.16532599...
>>> mlpr.train_log_.collect()
     ITERATION       ERROR
0            1   34.525655
1            2   82.656301
...
733        734   11.891081
734        735   11.891081

[735 rows x 2 columns]

>>> pred_df.collect()
   ID  V000  V001 V002  V003
0   1     1  1.71   AC     0
1   2    10  1.78   CA     5
2   3    17  2.36   AA     6

Invoke predict():

>>> res  = mlpr.predict(data=pred_df, key='ID')

Result may look different from the following results due to model randomness.

>>> res.collect()
   ID TARGET      VALUE
0   1   T001  12.700012
1   1   T002   2.799133
...
7   3   T002   2.799659
8   3   T003   2.190000

Attributes:

model_DataFrame

Model content.

train_log_DataFrame

Provides mean squared error between predicted values and target values for each iteration.

stats_DataFrame

Statistics.

optim_param_DataFrame

Provides optimal parameters selected.

Available only when parameter selection is triggered.

Methods

`create_model_state`([model, function, ...])	Create PAL model state.
`delete_model_state`([state])	Delete PAL model state.
`fit`(data[, key, features, label, ...])	Fit the model to the training dataset.
`predict`(data[, key, features, thread_ratio])	Predict dependent variable values based on a fitted model.
`score`(data, key[, features, label, thread_ratio])	Returns the coefficient of determination R2 of the prediction.
`set_model_state`(state)	Set the model state by state information.

fit(data, key=None, features=None, label=None, categorical_variable=None)

Fit the model to the training dataset.

Parameters:

dataDataFrame

DataFrame containing the data.

keystr, optional

Name of the ID column.

If key is not provided, then:

if data is indexed by a single column, then key defaults to that index column

otherwise, it is assumed that data contains no ID column

featuresa list of str, optional

Names of the feature columns.

If features is not provided, it defaults to all the non-ID and non-label columns.

labelstr or a list of str, optional

Name of the label column, or list of names of multiple label columns.

If label is not provided, it defaults to the last column.

categorical_variablestr or a list of str, optional

Specifies which INTEGER columns should be treated as categorical, with all other INTEGER columns treated as continuous.

No default value.

Returns:

A fitted object of class "MLPRegressor".

predict(data, key=None, features=None, thread_ratio=None)

Predict dependent variable values based on a fitted model.

Parameters:

dataDataFrame

DataFrame containing the data.

keystr, optional

Name of the ID column.

Mandatory if data is not indexed, or the index of data contains multiple columns.

Defaults to the single index column of data if not provided.

featuresa list of str, optional

Names of the feature columns.

If features is not provided, it defaults to all the non-ID columns.

thread_ratiofloat, optional

Adjusts the percentage of available threads to use, from 0 to 1. A value of 0 indicates the use of a single thread, while 1 implies the use of all possible current threads. Values outside the range will be ignored and this function heuristically determines the number of threads to use.

Defaults to 0.

Returns:

DataFrame

Predicted results, structured as follows:

ID column, with the same name and type as data 's ID column.

TARGET, type NVARCHAR, target name.

VALUE, type DOUBLE, regression value.

score(data, key, features=None, label=None, thread_ratio=None)

Returns the coefficient of determination R2 of the prediction.

Parameters:

dataDataFrame

DataFrame containing the data.

keystr, optional

Name of the ID column.

Mandatory if data is not indexed, or the index of data contains multiple columns.

Defaults to the single index column of data if not provided.

featuresa list of str, optional

Names of the feature columns.

If features is not provided, it defaults to all the non-ID and non-label columns.

labelstr or a list of str, optional

Name of the label column, or list of names of multiple label columns.

If label is not provided, it defaults to the last column.

Returns:

float: Returns the coefficient of determination R2 of the prediction.

create_model_state(model=None, function=None, pal_funcname='PAL_MULTILAYER_PERCEPTRON', state_description=None, force=False)

Create PAL model state.

Parameters:

modelDataFrame, optional

Specify the model for AFL state.

Defaults to self.model_.

functionstr, optional

Specify the function in the unified API.

A placeholder parameter, not effective for Multilayer Perceptron.

pal_funcnameint or str, optional

PAL function name.

Defaults to 'PAL_MULTILAYER_PERCEPTRON'.

state_descriptionstr, optional

Description of the state as model container.

Defaults to None.

forcebool, optional

If True it will delete the existing state.

Defaults to False.

delete_model_state(state=None)

Delete PAL model state.

Parameters:

stateDataFrame, optional

Specified the state.

Defaults to self.state.

set_model_state(state)

Set the model state by state information.

Parameters:

state: DataFrame or dict

If state is DataFrame, it has the following structure:

NAME: VARCHAR(100), it mush have STATE_ID, HINT, HOST and PORT.

VALUE: VARCHAR(1000), the values according to NAME.

If state is dict, the key must have STATE_ID, HINT, HOST and PORT.

Inherited Methods from PALBase

Besides those methods mentioned above, the MLPRegressor class also inherits methods from PALBase class, please refer to PAL Base for more details.