Access "sample_weight" in adaboost

[pierthodo]

Hello,

For a research paper I need to access the weights calculated by the Adaboostclassifier. Would it be possible to add the variable "sample_weight" as an attribute so I could publish code working with scikit-learn out of the box ?

Thank you,

Pierre

[jmschrei]

I'm not entirely sure how this would fit in with our current API. If you have a suggestion, the best idea is probably for you to submit a pull request and we'd be happy to review it.

[tgsmith61591]

From this block:

for iboost in range(self.n_estimators):
    # Boosting step
    sample_weight, estimator_weight, estimator_error = self._boost(
        iboost,
        X, y,
        sample_weight,
        random_state)

    # Early termination
    if sample_weight is None:
        break

    self.estimator_weights_[iboost] = estimator_weight
    self.estimator_errors_[iboost] = estimator_error

Wouldn't it be as simple as maintaining another list (similar to estimator_weights_)? For example (larger block for context):

# Clear any previous fit results
self.estimators_ = []
self.sample_weights_ = []  # add this list
self.estimator_weights_ = np.zeros(self.n_estimators, dtype=np.float64)
self.estimator_errors_ = np.ones(self.n_estimators, dtype=np.float64)

random_state = check_random_state(self.random_state)

for iboost in range(self.n_estimators):
    # Boosting step
    sample_weight, estimator_weight, estimator_error = self._boost(
        iboost,
        X, y,
        sample_weight,
        random_state)

    # Early termination
    if sample_weight is None:
        break

    self.sample_weights_.append(sample_weight)  # add this append statement
    self.estimator_weights_[iboost] = estimator_weight
    self.estimator_errors_[iboost] = estimator_error

(Happy to add and submit a PR, if you think it might be worthwhile)