Periodic Machine Learning crosses algorithms to improve the Artificial Intelligence models – Science

Like the Periodic Chemical Table, the Massachusetts Institute of Technology (MIT) has created a table dedicated to machine practice. The The goal is to create unified infrastructure to help improve AI models or to create new ones to combine existing ideas.

A. The periodic table of machine learning shows how more than 20 classic machine learning algorithms are connected. Researchers can create new strategies or if possible, create new solutions to merge different methods to improve existing AI models. It is possible to create a new image classification algorithm by combining the elements of two different algorithms, which has a better performance of 8% than the current GAMA solutions.

According to MIT, The table expands from an important idea: all algorithms learn a specific type of relationship between two data points. Each algorithm improves in a slightly different directionNuclear maths are the same behind each approach. On this basis, scientists have identified a unified equation based on several classic AI algorithms.

This equation has been used to reform popular methods and align them on the tableAll are classified based on the close relationships you have learned.

When a table of chemical elements was created, scientists left the space filled with innovations. The same applies to a new case dedicated to machine practice, ts that should exist where algorithms should exist, but they have not yet been found. “We have begun to look at the machine practice as a system with a structure that can explore the progress rather than ingesting in,” researcher Shaden Alshammeri Do Mitt said.

A. Table was inspired by a combination of algorithms in which the researcher invented the contact points of training data. And from here he tried to reflect others, classifying these algorithms. As the table was aligned, researchers began to observe the shortcomings of algorithms.

Can contact a full study Publication MIT MIT.