- Scikit Learn Tutorial
- Scikit Learn - Home
- Scikit Learn - Introduction
- Scikit Learn - Modelling Process
- Scikit Learn - Data Representation
- Scikit Learn - Estimator API
- Scikit Learn - Conventions
- Scikit Learn - Linear Modeling
- Scikit Learn - Extended Linear Modeling
- Stochastic Gradient Descent
- Scikit Learn - Support Vector Machines
- Scikit Learn - Anomaly Detection
- Scikit Learn - K-Nearest Neighbors
- Scikit Learn - KNN Learning
- Classification with Naïve Bayes
- Scikit Learn - Decision Trees
- Randomized Decision Trees
- Scikit Learn - Boosting Methods
- Scikit Learn - Clustering Methods
- Clustering Performance Evaluation
- Dimensionality Reduction using PCA
- Scikit Learn Useful Resources
- Scikit Learn - Quick Guide
- Scikit Learn - Useful Resources
- Scikit Learn - Discussion
- Selected Reading
- UPSC IAS Exams Notes
- Developer's Best Practices
- Questions and Answers
- Effective Resume Writing
- HR Interview Questions
- Computer Glossary
- Who is Who
Scikit Learn - Conventions
Scikit-learn’s objects share a uniform basic API that consists of the following three complementary interfaces −
Estimator interface − It is for building and fitting the models.
Predictor interface − It is for making predictions.
Transformer interface − It is for converting data.
The APIs adopt simple conventions and the design choices have been guided in a manner to avoid the proliferation of framework code.
Purpose of Conventions
The purpose of conventions is to make sure that the API stick to the following broad principles −
Consistency − All the objects whether they are basic, or composite must share a consistent interface which further composed of a limited set of methods.
Inspection − Constructor parameters and parameters values determined by learning algorithm should be stored and exposed as public attributes.
Non-proliferation of classes − Datasets should be represented as NumPy arrays or Scipy sparse matrix whereas hyper-parameters names and values should be represented as standard Python strings to avoid the proliferation of framework code.
Composition − The algorithms whether they are expressible as sequences or combinations of transformations to the data or naturally viewed as meta-algorithms parameterized on other algorithms, should be implemented and composed from existing building blocks.
Sensible defaults − In scikit-learn whenever an operation requires a user-defined parameter, an appropriate default value is defined. This default value should cause the operation to be performed in a sensible way, for example, giving a base-line solution for the task at hand.
The conventions available in Sklearn are explained below −
It states that the input should be cast to float64. In the following example, in which sklearn.random_projection module used to reduce the dimensionality of the data, will explain it −
import numpy as np from sklearn import random_projection rannge = np.random.RandomState(0) X = range.rand(10,2000) X = np.array(X, dtype = 'float32') X.dtype Transformer_data = random_projection.GaussianRandomProjection() X_new = transformer.fit_transform(X) X_new.dtype
In the above example, we can see that X is float32 which is cast to float64 by fit_transform(X).
Refitting & Updating Parameters
Hyper-parameters of an estimator can be updated and refitted after it has been constructed via the set_params() method. Let’s see the following example to understand it −
import numpy as np from sklearn.datasets import load_iris from sklearn.svm import SVC X, y = load_iris(return_X_y = True) clf = SVC() clf.set_params(kernel = 'linear').fit(X, y) clf.predict(X[:5])
array([0, 0, 0, 0, 0])
Once the estimator has been constructed, above code will change the default kernel rbf to linear via SVC.set_params().
Now, the following code will change back the kernel to rbf to refit the estimator and to make a second prediction.
clf.set_params(kernel = 'rbf', gamma = 'scale').fit(X, y) clf.predict(X[:5])
array([0, 0, 0, 0, 0])
The following is the complete executable program −
import numpy as np from sklearn.datasets import load_iris from sklearn.svm import SVC X, y = load_iris(return_X_y = True) clf = SVC() clf.set_params(kernel = 'linear').fit(X, y) clf.predict(X[:5]) clf.set_params(kernel = 'rbf', gamma = 'scale').fit(X, y) clf.predict(X[:5])
Multiclass & Multilabel fitting
In case of multiclass fitting, both learning and the prediction tasks are dependent on the format of the target data fit upon. The module used is sklearn.multiclass. Check the example below, where multiclass classifier is fit on a 1d array.
from sklearn.svm import SVC from sklearn.multiclass import OneVsRestClassifier from sklearn.preprocessing import LabelBinarizer X = [[1, 2], [3, 4], [4, 5], [5, 2], [1, 1]] y = [0, 0, 1, 1, 2] classif = OneVsRestClassifier(estimator = SVC(gamma = 'scale',random_state = 0)) classif.fit(X, y).predict(X)
array([0, 0, 1, 1, 2])
In the above example, classifier is fit on one dimensional array of multiclass labels and the predict() method hence provides corresponding multiclass prediction. But on the other hand, it is also possible to fit upon a two-dimensional array of binary label indicators as follows −
from sklearn.svm import SVC from sklearn.multiclass import OneVsRestClassifier from sklearn.preprocessing import LabelBinarizer X = [[1, 2], [3, 4], [4, 5], [5, 2], [1, 1]] y = LabelBinarizer().fit_transform(y) classif.fit(X, y).predict(X)
array( [ [0, 0, 0], [0, 0, 0], [0, 1, 0], [0, 1, 0], [0, 0, 0] ] )
Similarly, in case of multilabel fitting, an instance can be assigned multiple labels as follows −
from sklearn.preprocessing import MultiLabelBinarizer y = [[0, 1], [0, 2], [1, 3], [0, 2, 3], [2, 4]] y = MultiLabelBinarizer().fit_transform(y) classif.fit(X, y).predict(X)
array( [ [1, 0, 1, 0, 0], [1, 0, 1, 0, 0], [1, 0, 1, 1, 0], [1, 0, 1, 1, 0], [1, 0, 1, 0, 0] ] )
In the above example, sklearn.MultiLabelBinarizer is used to binarize the two dimensional array of multilabels to fit upon. That’s why predict() function gives a 2d array as output with multiple labels for each instance.