GSgnnRconstructFeatRegScoreEvaluator

class graphstorm.eval.GSgnnRconstructFeatRegScoreEvaluator(eval_frequency, eval_metric_list=None, use_early_stop=False, early_stop_burnin_rounds=0, early_stop_rounds=3, early_stop_strategy='average_increase')

Bases: GSgnnRegressionEvaluator

Evaluator for feature reconstruction tasks using regression scores.

A built-in evaluator for feature reconstruction tasks. It uses mse or rmse as evaluation metrics.

This evaluator requires the prediction results to be a 2D float tensor and the label also to be a 2D float tensor, which stores the original features.

Parameters

eval_frequency: int

The frequency (number of iterations) of doing evaluation.

eval_metric_list: list of string

Evaluation metrics used during evaluation. Default: [“mse”].

use_early_stop: bool

Set true to use early stop. Default: False.

early_stop_burnin_rounds: int

Burn-in rounds (number of evaluations) before starting to check for the early stop condition. Default: 0.

early_stop_rounds: int

The number of rounds (number of evaluations) for validation scores used to decide early stop. Default: 3.

early_stop_strategy: str

The early stop strategy. GraphStorm supports two strategies: 1) consecutive_increase, and 2) average_increase. Default: average_increase.

compute_score(pred, labels, train=True)

Compute feature reconstruction evaluation scores.

Parameters

pred: 2D tensor

The 2D tensor stores the prediction results.

labels: 2D tensor

The 2D tensor stores the labels that are the original node features as this is a feature reconstruction task.

train: bool

If in model training.

Returns

scores: dict

Evaluation scores of different feature reconstruction metrics in the format of {metric: score}. If either pred or labels are None, the score will be “N/A”.