AI Science and Technology Review Press: "There is no free lunch theorem" was once one of the most frequently talked about theorems in the field of machine learning (naturally, the more data, the better performance). However, Andreas Mueller, a machine learning scientist, recently wrote that everyone cited the wrong theorem. In fact, things are more complicated and far-reaching than this.

Andreas Mueller is an assistant research scientist at the Institute of Data Science of Columbia University and the author of Introduction to Python Machine Learning. He is also one of the core developers of scikit-learn machine learning library.

AI Technology Review compiled the full text of his blog as follows.

First of all, I want to sum up what I want to say in this article in one sentence: Try not to quote Wolpert's "No Free Lunch Theorem" in the future. If you quote it somewhere, you are likely to use it to support the wrong conclusion. What he really means is "you can't learn from data without assumptions".

The concept of "no free lunch theorem" was actually put forward by David H. Wolpert's "Lack of transcendental differences between learning algorithms" (/2019/07/don-cite-no-free-lunch-theory.html).

20 19 global summit on artificial intelligence and robotics

20 19 July 12 to 14, 2019 Global Summit on Artificial Intelligence and Robotics (hereinafter referred to as CCF-GAIR 20 19) was hosted by China Computer Federation and assisted by Lei Feng. Com and the Chinese University of Hong Kong (Shenzhen).

James II, the then Nobel Prize winner. Heckman, Chinese and foreign academicians, President of the World Summit, well-known Fellow and many important guests will come to the site to discuss the complex living conditions of artificial intelligence and robotics, such as learning, production and investment.