MLPRegressor

Defined in: ds/src/ml/estimators/MLPRegressor.js:19

Extends

Regressor

Constructors

Constructor

new MLPRegressor(params?): MLPRegressor

Defined in: ds/src/ml/estimators/MLPRegressor.js:20

Parameters

params?

Returns

MLPRegressor

Overrides

Regressor.constructor

Properties

_state

_state: object

Defined in: ds/src/core/estimators/estimator.js:27

Inherited from

Regressor._state

_warnings

_warnings: any[]

Defined in: ds/src/core/estimators/estimator.js:29

Inherited from

Regressor._warnings

fitted

fitted: boolean

Defined in: ds/src/core/estimators/estimator.js:25

Inherited from

Regressor.fitted

model

model: any

Defined in: ds/src/ml/estimators/MLPRegressor.js:24

params

params: object

Defined in: ds/src/ml/estimators/MLPRegressor.js:23

activation

activation: string = 'relu'

batchSize

batchSize: number = 32

epochs

epochs: number = 100

layerSizes

layerSizes: any = null

learningRate

learningRate: number = 0.01

omit_missing

omit_missing: boolean = true

verbose

verbose: boolean = false

Inherited from

Regressor.params

Methods

_prepareArgsForFit()

_prepareArgsForFit(args?): { columns?: undefined; columnsX: any[]; prepared: boolean; raw?: undefined; rows: any[]; X: any[][]; y: any[]; } | { columns: any[]; columnsX?: undefined; prepared: boolean; raw?: undefined; rows: any[]; X: any[][]; y?: undefined; } | { columns?: undefined; columnsX?: undefined; prepared?: undefined; raw: any[]; rows?: undefined; X?: undefined; y?: undefined; }

Defined in: ds/src/core/estimators/estimator.js:367

Convenience helper: parse arguments passed to fit/predict/transform.

Supports declarative table-style inputs:

fit({ X, y, data, omit_missing })
fit({ data, columns, … })

Returns an object { X, y, prepared, rows } where X/y are numeric arrays if preparation was required, otherwise returns the original values.

Note: this helper only prepares numeric matrices/vectors using core table utilities; it does not perform encoding of categorical predictors.

Parameters

args?

any[] = []

Returns

{ columns?: undefined; columnsX: any[]; prepared: boolean; raw?: undefined; rows: any[]; X: any[][]; y: any[]; } | { columns: any[]; columnsX?: undefined; prepared: boolean; raw?: undefined; rows: any[]; X: any[][]; y?: undefined; } | { columns?: undefined; columnsX?: undefined; prepared?: undefined; raw: any[]; rows?: undefined; X?: undefined; y?: undefined; }

Inherited from

Regressor._prepareArgsForFit

_r2()

_r2(yTrue, yPred): number

Defined in: ds/src/core/estimators/estimator.js:489

Parameters

yTrue

any

yPred

any

Returns

number

Inherited from

Regressor._r2

_repr_html_()

_repr_html_(): string

Defined in: ds/src/core/estimators/estimator.js:201

Observable/Jupyter HTML representation

Returns

string

HTML representation

Inherited from

Regressor._repr_html_

clearWarnings()

clearWarnings(): void

Defined in: ds/src/core/estimators/estimator.js:139

Clear all warnings

Returns

void

Inherited from

Regressor.clearWarnings

evaluate()

evaluate(X, y): any

Defined in: ds/src/ml/estimators/MLPRegressor.js:129

Parameters

X

any

y

any

Returns

any

fit()

fit(X, y?, opts?): MLPRegressor

Defined in: ds/src/ml/estimators/MLPRegressor.js:41

Fit the multilayer perceptron regressor on training data.

Parameters

X

any

Feature matrix (n samples × p features), or a declarative spec { X, columns, y, data, omit_missing }

y?

number[] = null

Target values (ignored when X is a spec)

opts?

Training hyperparameter overrides

activation?

string

Activation function name

batchSize?

number

Mini-batch size

epochs?

number

Number of training epochs

layerSizes?

number[]

Hidden/output layer sizes

learningRate?

number

Learning rate

verbose?

boolean

Log training progress

Returns

MLPRegressor

The fitted estimator (for chaining)

Overrides

Regressor.fit

getMemoryUsage()

getMemoryUsage(): string

Defined in: ds/src/core/estimators/estimator.js:97

Get memory usage in human-readable format

Returns

string

Memory usage string (e.g., “2.3 MB” or “145 KB”)

Inherited from

Regressor.getMemoryUsage

getParams()

getParams(): any

Defined in: ds/src/core/estimators/estimator.js:294

Get a shallow copy of parameters.

Returns

any

Inherited from

Regressor.getParams

getState()

getState(): any

Defined in: ds/src/core/estimators/estimator.js:65

Get comprehensive model state

Returns

any

State information including fitted status, memory estimate, warnings

Inherited from

Regressor.getState

getWarnings()

getWarnings(): any[]

Defined in: ds/src/core/estimators/estimator.js:124

Get all warnings

Returns

any[]

Array of warning objects

Inherited from

Regressor.getWarnings

getWarningsByType()

getWarningsByType(type): any[]

Defined in: ds/src/core/estimators/estimator.js:148

Get warnings of a specific type

Parameters

type

string

Warning type

Returns

any[]

Filtered warnings

Inherited from

Regressor.getWarningsByType

hasWarnings()

hasWarnings(): boolean

Defined in: ds/src/core/estimators/estimator.js:132

Check if model has warnings

Returns

boolean

Inherited from

Regressor.hasWarnings

isFitted()

isFitted(): boolean

Defined in: ds/src/core/estimators/estimator.js:36

Check if model is fitted

Returns

boolean

Inherited from

Regressor.isFitted

predict()

predict(X): number[]

Defined in: ds/src/ml/estimators/MLPRegressor.js:104

Predict target values for samples in X.

Parameters

X

any

Feature matrix (n samples × p features), or a declarative spec { X, columns, data, omit_missing }

Returns

number[]

Predicted target values

Overrides

Regressor.predict

save()

save(): string

Defined in: ds/src/core/estimators/estimator.js:329

Save model to JSON string

Returns

string

JSON representation of the model

Inherited from

Regressor.save

score()

score(yTrueOrOpts, yPred, _opts?, …args?): number

Defined in: ds/src/core/estimators/estimator.js:461

Default R^2 scoring implementation: 1 - SS_res / SS_tot

Accepts either:

arrays: score(yTrue, yPred)
table-style: score({ X, y, data }) where predict will be called internally

Parameters

yTrueOrOpts

any

yPred

any

_opts?

args?

…any[] = {}

Returns

number

Inherited from

Regressor.score

setParams()

setParams(params?): MLPRegressor

Defined in: ds/src/core/estimators/estimator.js:285

Set parameters (mutates instance).

Parameters

params?

any = {}

Returns

MLPRegressor

Inherited from

Regressor.setParams

summary()

summary(): object

Defined in: ds/src/ml/estimators/MLPRegressor.js:136

Returns

object

epochs

epochs: any

finalLoss

finalLoss: any

initialLoss

initialLoss: any

layerSizes

layerSizes: any

losses

losses: any

toJSON()

toJSON(): object

Defined in: ds/src/ml/estimators/MLPRegressor.js:150

Serialize minimal model metadata. Subclasses may override to include learned parameters.

Returns

object

class

__class__: string = 'MLPRegressor'

fitted

fitted: boolean

model

model: any

params

params: any

Overrides

Regressor.toJSON

transform()

transform(): void

Defined in: ds/src/core/estimators/estimator.js:431

Transform should be implemented by transformers.

Returns

void

Inherited from

Regressor.transform

fromJSON()

static fromJSON(obj?): MLPRegressor

Defined in: ds/src/ml/estimators/MLPRegressor.js:159

Basic deserialization. Subclasses should override if they need to restore learned arrays / matrices.

Parameters

obj?

Returns

MLPRegressor

Overrides

Regressor.fromJSON

load()

static load(jsonString): Estimator

Defined in: ds/src/core/estimators/estimator.js:346

Load model from JSON string

Parameters

jsonString

string

JSON representation

Returns

Estimator

Reconstructed estimator instance

Inherited from

Regressor.load