SGDClassifier
Linear classifiers (SVM, logistic regression, etc.) with SGD training.
This estimator implements regularized linear models with stochastic gradient descent (SGD) learning: the gradient of the loss is estimated each sample at a time and the model is updated along the way with a decreasing strength schedule (aka learning rate). SGD allows minibatch (online/out-of-core) learning via the partial\_fit method. For best results using the default learning rate schedule, the data should have zero mean and unit variance.
This implementation works with data represented as dense or sparse arrays of floating point values for the features. The model it fits can be controlled with the loss parameter; by default, it fits a linear support vector machine (SVM).
The regularizer is a penalty added to the loss function that shrinks model parameters towards the zero vector using either the squared euclidean norm L2 or the absolute norm L1 or a combination of both (Elastic Net). If the parameter update crosses the 0.0 value because of the regularizer, the update is truncated to 0.0 to allow for learning sparse models and achieve online feature selection.
Read more in the User Guide.
Python Reference (opens in a new tab)
Constructors
constructor()
Signature
new SGDClassifier(opts?: object): SGDClassifier;Parameters
| Name | Type | Description |
|---|---|---|
opts? | object | - |
opts.alpha? | number | Constant that multiplies the regularization term. The higher the value, the stronger the regularization. Also used to compute the learning rate when learning\_rate is set to ‘optimal’. Values must be in the range \[0.0, inf). Default Value 0.0001 |
opts.average? | number | boolean | When set to true, computes the averaged SGD weights across all updates and stores the result in the coef\_ attribute. If set to an int greater than 1, averaging will begin once the total number of samples seen reaches average. So average=10 will begin averaging after seeing 10 samples. Integer values must be in the range \[1, n\_samples\]. Default Value false |
opts.class_weight? | any | Preset for the class_weight fit parameter. Weights associated with classes. If not given, all classes are supposed to have weight one. The “balanced” mode uses the values of y to automatically adjust weights inversely proportional to class frequencies in the input data as n\_samples / (n\_classes \* np.bincount(y)). |
opts.early_stopping? | boolean | Whether to use early stopping to terminate training when validation score is not improving. If set to true, it will automatically set aside a stratified fraction of training data as validation and terminate training when validation score returned by the score method is not improving by at least tol for n_iter_no_change consecutive epochs. Default Value false |
opts.epsilon? | number | Epsilon in the epsilon-insensitive loss functions; only if loss is ‘huber’, ‘epsilon_insensitive’, or ‘squared_epsilon_insensitive’. For ‘huber’, determines the threshold at which it becomes less important to get the prediction exactly right. For epsilon-insensitive, any differences between the current prediction and the correct label are ignored if they are less than this threshold. Values must be in the range \[0.0, inf). Default Value 0.1 |
opts.eta0? | number | The initial learning rate for the ‘constant’, ‘invscaling’ or ‘adaptive’ schedules. The default value is 0.0 as eta0 is not used by the default schedule ‘optimal’. Values must be in the range (0.0, inf). Default Value 0 |
opts.fit_intercept? | boolean | Whether the intercept should be estimated or not. If false, the data is assumed to be already centered. Default Value true |
opts.l1_ratio? | number | The Elastic Net mixing parameter, with 0 <= l1_ratio <= 1. l1_ratio=0 corresponds to L2 penalty, l1_ratio=1 to L1. Only used if penalty is ‘elasticnet’. Values must be in the range \[0.0, 1.0\]. Default Value 0.15 |
opts.learning_rate? | string | The learning rate schedule: Default Value 'optimal' |
opts.loss? | "log" | "log_loss" | "squared_error" | "huber" | "hinge" | "modified_huber" | "squared_hinge" | "perceptron" | "epsilon_insensitive" | "squared_epsilon_insensitive" | The loss function to be used. Default Value 'hinge' |
opts.max_iter? | number | The maximum number of passes over the training data (aka epochs). It only impacts the behavior in the fit method, and not the partial\_fit method. Values must be in the range \[1, inf). Default Value 1000 |
opts.n_iter_no_change? | number | Number of iterations with no improvement to wait before stopping fitting. Convergence is checked against the training loss or the validation loss depending on the early\_stopping parameter. Integer values must be in the range \[1, max\_iter). Default Value 5 |
opts.n_jobs? | number | The number of CPUs to use to do the OVA (One Versus All, for multi-class problems) computation. undefined means 1 unless in a joblib.parallel\_backend (opens in a new tab) context. \-1 means using all processors. See Glossary for more details. |
opts.penalty? | "l1" | "l2" | "elasticnet" | The penalty (aka regularization term) to be used. Defaults to ‘l2’ which is the standard regularizer for linear SVM models. ‘l1’ and ‘elasticnet’ might bring sparsity to the model (feature selection) not achievable with ‘l2’. No penalty is added when set to undefined. Default Value 'l2' |
opts.power_t? | number | The exponent for inverse scaling learning rate [default 0.5]. Values must be in the range (-inf, inf). Default Value 0.5 |
opts.random_state? | number | Used for shuffling the data, when shuffle is set to true. Pass an int for reproducible output across multiple function calls. See Glossary. Integer values must be in the range \[0, 2\*\*32 \- 1\]. |
opts.shuffle? | boolean | Whether or not the training data should be shuffled after each epoch. Default Value true |
opts.tol? | number | The stopping criterion. If it is not undefined, training will stop when (loss > best_loss - tol) for n\_iter\_no\_change consecutive epochs. Convergence is checked against the training loss or the validation loss depending on the early\_stopping parameter. Values must be in the range \0.0, inf). Default Value 0.001 |
opts.validation_fraction? | number | The proportion of training data to set aside as validation set for early stopping. Must be between 0 and 1. Only used if early\_stopping is true. Values must be in the range (0.0, 1.0). Default Value 0.1 |
opts.verbose? | number | The verbosity level. Values must be in the range \[0, inf). Default Value 0 |
opts.warm_start? | boolean | When set to true, reuse the solution of the previous call to fit as initialization, otherwise, just erase the previous solution. See [the Glossary. Repeatedly calling fit or partial_fit when warm_start is true can result in a different solution than when calling fit a single time because of the way the data is shuffled. If a dynamic learning rate is used, the learning rate is adapted depending on the number of samples already seen. Calling fit resets this counter, while partial\_fit will result in increasing the existing counter. Default Value false |
Returns
Defined in: generated/linear_model/SGDClassifier.ts:29 (opens in a new tab)
Properties
_isDisposed
boolean=false
Defined in: generated/linear_model/SGDClassifier.ts:27 (opens in a new tab)
_isInitialized
boolean=false
Defined in: generated/linear_model/SGDClassifier.ts:26 (opens in a new tab)
_py
PythonBridge
Defined in: generated/linear_model/SGDClassifier.ts:25 (opens in a new tab)
id
string
Defined in: generated/linear_model/SGDClassifier.ts:22 (opens in a new tab)
opts
any
Defined in: generated/linear_model/SGDClassifier.ts:23 (opens in a new tab)
Accessors
classes_
Signature
classes_(): Promise<any[]>;Returns
Promise<any[]>
Defined in: generated/linear_model/SGDClassifier.ts:747 (opens in a new tab)
coef_
Weights assigned to the features.
Signature
coef_(): Promise<ArrayLike[][]>;Returns
Promise<ArrayLike[][]>
Defined in: generated/linear_model/SGDClassifier.ts:657 (opens in a new tab)
feature_names_in_
Names of features seen during fit. Defined only when X has feature names that are all strings.
Signature
feature_names_in_(): Promise<ArrayLike>;Returns
Promise<ArrayLike>
Defined in: generated/linear_model/SGDClassifier.ts:820 (opens in a new tab)
intercept_
Constants in decision function.
Signature
intercept_(): Promise<ArrayLike[]>;Returns
Promise<ArrayLike[]>
Defined in: generated/linear_model/SGDClassifier.ts:680 (opens in a new tab)
loss_function_
Signature
loss_function_(): Promise<any>;Returns
Promise<any>
Defined in: generated/linear_model/SGDClassifier.ts:725 (opens in a new tab)
n_features_in_
Number of features seen during fit.
Signature
n_features_in_(): Promise<number>;Returns
Promise<number>
Defined in: generated/linear_model/SGDClassifier.ts:795 (opens in a new tab)
n_iter_
The actual number of iterations before reaching the stopping criterion. For multiclass fits, it is the maximum over every binary fit.
Signature
n_iter_(): Promise<number>;Returns
Promise<number>
Defined in: generated/linear_model/SGDClassifier.ts:705 (opens in a new tab)
py
Signature
py(): PythonBridge;Returns
PythonBridge
Defined in: generated/linear_model/SGDClassifier.ts:191 (opens in a new tab)
Signature
py(pythonBridge: PythonBridge): void;Parameters
| Name | Type |
|---|---|
pythonBridge | PythonBridge |
Returns
void
Defined in: generated/linear_model/SGDClassifier.ts:195 (opens in a new tab)
t_
Number of weight updates performed during training. Same as (n\_iter\_ \* n\_samples + 1).
Signature
t_(): Promise<number>;Returns
Promise<number>
Defined in: generated/linear_model/SGDClassifier.ts:772 (opens in a new tab)
Methods
decision_function()
Predict confidence scores for samples.
The confidence score for a sample is proportional to the signed distance of that sample to the hyperplane.
Signature
decision_function(opts: object): Promise<ArrayLike>;Parameters
| Name | Type | Description |
|---|---|---|
opts | object | - |
opts.X? | ArrayLike | The data matrix for which we want to get the confidence scores. |
Returns
Promise<ArrayLike>
Defined in: generated/linear_model/SGDClassifier.ts:287 (opens in a new tab)
densify()
Convert coefficient matrix to dense array format.
Converts the coef\_ member (back) to a numpy.ndarray. This is the default format of coef\_ and is required for fitting, so calling this method is only required on models that have previously been sparsified; otherwise, it is a no-op.
Signature
densify(opts: object): Promise<any>;Parameters
| Name | Type |
|---|---|
opts | object |
Returns
Promise<any>
Defined in: generated/linear_model/SGDClassifier.ts:324 (opens in a new tab)
dispose()
Disposes of the underlying Python resources.
Once dispose() is called, the instance is no longer usable.
Signature
dispose(): Promise<void>;Returns
Promise<void>
Defined in: generated/linear_model/SGDClassifier.ts:268 (opens in a new tab)
fit()
Fit linear model with Stochastic Gradient Descent.
Signature
fit(opts: object): Promise<any>;Parameters
| Name | Type | Description |
|---|---|---|
opts | object | - |
opts.X? | any | Training data. |
opts.coef_init? | ArrayLike[] | The initial coefficients to warm-start the optimization. |
opts.intercept_init? | ArrayLike | The initial intercept to warm-start the optimization. |
opts.sample_weight? | ArrayLike | Weights applied to individual samples. If not provided, uniform weights are assumed. These weights will be multiplied with class_weight (passed through the constructor) if class_weight is specified. |
opts.y? | ArrayLike | Target values. |
Returns
Promise<any>
Defined in: generated/linear_model/SGDClassifier.ts:350 (opens in a new tab)
init()
Initializes the underlying Python resources.
This instance is not usable until the Promise returned by init() resolves.
Signature
init(py: PythonBridge): Promise<void>;Parameters
| Name | Type |
|---|---|
py | PythonBridge |
Returns
Promise<void>
Defined in: generated/linear_model/SGDClassifier.ts:204 (opens in a new tab)
partial_fit()
Perform one epoch of stochastic gradient descent on given samples.
Internally, this method uses max\_iter \= 1. Therefore, it is not guaranteed that a minimum of the cost function is reached after calling it once. Matters such as objective convergence, early stopping, and learning rate adjustments should be handled by the user.
Signature
partial_fit(opts: object): Promise<any>;Parameters
| Name | Type | Description |
|---|---|---|
opts | object | - |
opts.X? | any | Subset of the training data. |
opts.classes? | ArrayLike | Classes across all calls to partial_fit. Can be obtained by via np.unique(y\_all), where y_all is the target vector of the entire dataset. This argument is required for the first call to partial_fit and can be omitted in the subsequent calls. Note that y doesn’t need to contain all labels in classes. |
opts.sample_weight? | ArrayLike | Weights applied to individual samples. If not provided, uniform weights are assumed. |
opts.y? | ArrayLike | Subset of the target values. |
Returns
Promise<any>
Defined in: generated/linear_model/SGDClassifier.ts:413 (opens in a new tab)
predict()
Predict class labels for samples in X.
Signature
predict(opts: object): Promise<ArrayLike>;Parameters
| Name | Type | Description |
|---|---|---|
opts | object | - |
opts.X? | ArrayLike | The data matrix for which we want to get the predictions. |
Returns
Promise<ArrayLike>
Defined in: generated/linear_model/SGDClassifier.ts:465 (opens in a new tab)
predict_log_proba()
Log of probability estimates.
This method is only available for log loss and modified Huber loss.
When loss=”modified_huber”, probability estimates may be hard zeros and ones, so taking the logarithm is not possible.
See predict\_proba for details.
Signature
predict_log_proba(opts: object): Promise<ArrayLike>;Parameters
| Name | Type | Description |
|---|---|---|
opts | object | - |
opts.X? | ArrayLike | Input data for prediction. |
Returns
Promise<ArrayLike>
Defined in: generated/linear_model/SGDClassifier.ts:504 (opens in a new tab)
predict_proba()
Probability estimates.
This method is only available for log loss and modified Huber loss.
Multiclass probability estimates are derived from binary (one-vs.-rest) estimates by simple normalization, as recommended by Zadrozny and Elkan.
Binary probability estimates for loss=”modified_huber” are given by (clip(decision_function(X), -1, 1) + 1) / 2. For other loss functions it is necessary to perform proper probability calibration by wrapping the classifier with CalibratedClassifierCV instead.
Signature
predict_proba(opts: object): Promise<any>;Parameters
| Name | Type | Description |
|---|---|---|
opts | object | - |
opts.X? | any | Input data for prediction. |
Returns
Promise<any>
Defined in: generated/linear_model/SGDClassifier.ts:545 (opens in a new tab)
score()
Return the mean accuracy on the given test data and labels.
In multi-label classification, this is the subset accuracy which is a harsh metric since you require for each sample that each label set be correctly predicted.
Signature
score(opts: object): Promise<number>;Parameters
| Name | Type | Description |
|---|---|---|
opts | object | - |
opts.X? | ArrayLike[] | Test samples. |
opts.sample_weight? | ArrayLike | Sample weights. |
opts.y? | ArrayLike | True labels for X. |
Returns
Promise<number>
Defined in: generated/linear_model/SGDClassifier.ts:580 (opens in a new tab)
sparsify()
Convert coefficient matrix to sparse format.
Converts the coef\_ member to a scipy.sparse matrix, which for L1-regularized models can be much more memory- and storage-efficient than the usual numpy.ndarray representation.
The intercept\_ member is not converted.
Signature
sparsify(opts: object): Promise<any>;Parameters
| Name | Type |
|---|---|
opts | object |
Returns
Promise<any>
Defined in: generated/linear_model/SGDClassifier.ts:631 (opens in a new tab)