Is there a universal logic for the identification of cats and the operation of the universe jkforum

Is there a universal logic for the identification of cats and the operation of the universe? Source: Mr. WeChat public early in 2012, the computer will have learned to identify YouTube video of the cat; by November 2014, the computer can even be a photo correctly named as "a group of young people playing frisbee, and artificial intelligence researchers cheered for" deep learning "will bring more this achievement, the success of the algorithm to the brain thinking, that only through contact develop sensitivity to the characteristics of the real world. Deep learning through the use of the latest design, scientists have become increasingly adept at applying artificial neural network architecture model of computer application in the field of image, voice and pattern recognition and machine — they are personal assistants, complex data analysis and automatic driving vehicle related core technology. However, in addition to training the computer to extract some salient features from other irrelevant data, researchers have never fully understood why this algorithm, or biological learning method, can work. A study of two physicists found that deep learning in some form, its mechanism of action as a mathematical method in physics is the most important and the most common calculation method, a physical system of large scale behavior, it is commonly used in the basic particle, fluid calculation and cosmology. The findings from the Boston University pan cagy · Mehta (Pankaj Mehta) and David · of Northwestern University; Schwab (David Schwab) jointly completed. They proved a called "renormalization" (renormalization) technology can realize statistical data classification to artificial neural network, such as the identification of "a cat" in a given video, regardless of their color, size or position. This statistical method was originally used in the field of physics, which makes it possible for physicists to describe the large-scale system without knowing the exact state of all components. Pan cagy · Mehta, assistant professor of physics at Boston University (photo: Pan cagy · Mehta) "these had just dream, they use evidence in a paper," Emory University (Emory University) biological physicist Leah · in Mengman (Ilya Nemenman) says, "extracting relevant features in the field of statistical physics, and extract relevant features of deep learning in the field, not just the same argument, which in essence is one thing." We humans have a special knack for identifying a cat in a bush, a familiar face in the crowd, or an arbitrary target surrounded by color, texture, and sound. The strong similarity between the biology of the learning process and the depth of learning of the machine suggests that the brain has also adopted some form of renormalization to understand the world. "Extracting relevant features from the data, perhaps there is a general logic," Mehta said. "I think it’s a.相关的主题文章: