Scikit-learn随机森林冗长输出解读：Parallel日志中task的含义疑问

Great question! Let's break down what those log lines mean, especially the "task" part:

• First, to answer your core question directly: Yes, each "task" in those logs corresponds to training a single estimator (decision tree) in your Random Forest.

When you set n_estimators=100 in your RandomForestClassifier/Regressor, Scikit-learn will create 100 independent decision trees. Since each tree is built using a separate bootstrap sample of your data and a random subset of features, they can be trained in parallel. Each of these tree-training jobs counts as one "task" in the parallel processing log.

Looking at your example output:

[Parallel(n_jobs=-2)]: Done 50 out of 50 | elapsed: 1366.7min finished

This means your Random Forest was configured with n_estimators=50—all 50 trees (tasks) completed successfully.

Let's quickly cover the other parts of the log for full context:

• Parallel(n_jobs=-2): The n_jobs=-2 setting tells Scikit-learn to use all available CPU cores minus one (hence the 11 concurrent workers here, implying you have 12 total cores).
• Using backend ThreadingBackend with 11 concurrent workers: This specifies that parallel processing is done via threads (the standard backend for CPU-bound tree training tasks in Scikit-learn).
• The elapsed time tracks how long it took to complete the batch of tasks shown (28 tasks taking ~902 minutes, then all 50 taking ~1367 minutes total).

To sum up: Each task is exactly one decision tree being trained. So if you set n_estimators=100, you'll see logs tracking progress up to 100 tasks completed.

内容的提问来源于stack exchange，提问作者Ehsan Esbati