As well as MLPClassifier, there is its regressor sibling,MLPRegressor. The two share an almost identical interface.The main difference between the two is the loss functions used by each of them and the activation functions of the output layer. The regressor optimizes a squared loss, and the last layer is activated by an identity function. All other hyperparameters are the same, including the four activation options for the hidden layers.
Both estimators have a partial_fit()method. You can use it to update the model once you get a hold of additional training data after the estimator has already been fitted.score()in MLPRegressorcalculates the regressor'sR2, as opposed to the classifier's accuracy, which is calculated byMLPClassifier.
