当前位置：网站首页>What is ensemble learning in machine learning?

What is ensemble learning in machine learning?

2022-08-11 03:51:00 【Program Yuan Keke】

Ensemble learning is one of the most powerful machine learning techniques.Ensemble learning improves the reliability and accuracy of prediction results by using multiple machine learning models.But how does using multiple machine learning models make predictions more accurate? What techniques can be used to create an ensemble learning model? The following sections explore answering these questions and examine the rationale for using ensemble models and the main methods for creating ensemble models.

What is ensemble learning?

In short, ensemble learning is the process of training multiple machine learning models and combining their outputs.The organization strives to build an optimal forecasting model based on different models.Combining various machine learning models can improve the stability of the overall model, resulting in more accurate predictions.Ensemble learning models are generally more reliable than individual models, and as a result, they often win many machine learning competitions.

Engineers can use a variety of techniques to create ensemble learning models.While simple ensemble learning techniques involve averaging the outputs of different models, more complex methods and algorithms have been developed specifically to combine the predictions of many underlying learners/models.

Why use ensemble training?

Machine learning models can differ from one another for a number of reasons.Different machine learning models can operate on different samples of the population data, can use different modeling techniques, and use different assumptions.

Imagine if you join a team of different professionals, there must be some techniques you know and don't know, assuming you're discussing with other membersa technical topic.Like you, they only know something about their specialty and nothing else.However, if these technical knowledge can finally be combined, there will be more accurate guesses in more domains, which is the principle of ensemble learning, that is, combining predictions from different individual models (team members) to improve accuracy, andMinimize errors.

Statisticians have shown that when a group of people is asked to use a range of possible answers to guess the correct answer to a given question, all their answers form aProbability distributions.Those who really know the right answer will confidently choose the correct answer, while those who choose the wrong answer will spread their guesses across the range of possible wrong answers.For example, in a guessing game, if you and two friends both know the correct answer is A, then all three of you will choose A, while the other three people on the team who do not know the answer are likely to incorrectly guess B, C,D or E, the result is that A has three votes, other answers may only have one or two votes.

All models have certain errors.The error of one model will be different from the error produced by another model because the models themselves are different for the reasons mentioned above.When all errors are checked, they are not clustered around a single answer, but widely distributed.Incorrect guesses are basically spread over all possible wrong answers and cancel each other out.At the same time, correct guesses from different models will cluster around the correct answer.More reliable correct answers can be found when using ensemble training methods.

Free to share some artificial intelligence learning materials that I have organized for everyone. It has been organized for a long time and is very comprehensive.Including some basic introduction videos of artificial intelligence + practical videos of common AI frameworks, image recognition, OpenCV, NLP, YOLO, machine learning, pytorch, computer vision, deep learning and neural networks and other videos, courseware source code, domestic and foreign well-known essence resources, AI popularpapers, etc.
The following are some screenshots, and the free download method is attached at the end of the article.

Table of Contents