When I am alone, I feel a little lonely, what should I do? "Microsoft Xiao Bing" launched by Microsoft Research Asia may be able to chat with you like a girlfriend. In addition to the functions of “appreciation value†and “choice matchingâ€, the 3.0 version of “Xiaobing†also has powerful visual recognition capabilities based on deep learning technology. After seeing a picture, it can give a humanized response based on emotion, and the second back speed is shortened to 250 milliseconds.
Not only "Microsoft Little Ice" and Go Master "Alpha Dog", from Internet search to language translation, to even identifying genes at risk of autism... Any area that needs to predict unknown information from large amounts of data is deep Learn where you can get a fist. So what is deep learning technology? How will it change the lives of human beings?
Duplicate cat found in 10,000 pictures
In 2011, researchers at a Google lab took 10 million still images from a video site and “fed†it to Google’s brain, with the goal of finding repetitive patterns. Three days later, Google’s brain discovered “cats†from these pictures without human help.
This Google brain is a large neural network model using deep learning technology, consisting of 1000 computers. This incident caused a sensation in the scientific and technological community and was considered a milestone in the revival of deep learning.
The so-called deep learning is a neural network composed of multi-layered neurons to achieve the function of machine learning. These multi-layered computer networks, like the human brain, collect information and generate behavior based on the information collected.
Traditional machine learning methods generally only exploit simple linear relationships, such as 1+1 equals 2. However, the world is not described by such a simple relationship, such as the relationship between income and age, gender, occupation, and education. The emergence of deep learning has changed this situation, and its inspiration comes from imitating the neural network of the human brain.
Scientists have found that the human cerebral cortex does not directly extract the features of the data transmitted by the retina, but allows the received stimulus signals to be screened through a complex network model. This hierarchical structure greatly reduces the amount of data processed by the vision system and ultimately retains useful information.
In the 1960s, when biologists studied the cat's cortex, they found that their unique network structure can effectively reduce the complexity of the feedback neural network, and then proposed a "convolution neural network." The deep learning program written by using this network structure has strong adaptability and becomes a breakthrough for artificial intelligence.
Speech recognition changes human-computer interaction
Simply put, deep learning technology is a simulation of the human brain, so it can accomplish the functions of many human brains.
The most well known is the visual function. Our camera can see the world like an eye, but we can't understand the world like a brain. Deep learning just fills this short board. With deep learning, Baidu maps can accurately identify the categories of objects in a photo and automatically categorize or search for photos. With deep learning, we can easily pay for your face. With deep learning, special machines can detect the movements of all people and vehicles in a certain space, and promptly report alarms for suspicious and dangerous events.
At the same time, deep learning technology has a wide range of applications in speech recognition. With the help of deep learning, computers have more and more powerful speech recognition capabilities, which may change the keyboard-based human-computer interaction mode.
Deep learning is also combined with enhanced learning, which is profoundly changing the field of robotics. The so-called enhanced learning refers to the strategy of self-learning better by the reward and punishment that the robot obtains through interaction with the environment. The "Alpha Dog" that attracted attention some time ago is the product of enhanced learning. It explores a better chess strategy by playing chess with the players or winning or losing with them.
What makes deep learning achieve beyond
However, creating a powerful neural network requires more processing layers. Due to hardware limitations, only 2 to 3 neural layers can be manufactured in the early stage. So, what makes deep learning beyond?
Obviously, the improvement of high-performance computing power is a big boost. These years of rapid development of GPUs, supercomputers, and cloud computing have made deep learning stand out. In 2011, Google brain used about 1,000 machines and 16,000 CPUs to process a deep learning model with about 1 billion neurons. Now, we can already do the same calculation on several GPUs.
"Deep learning is also helped by big data, just like the rocket has fuel." Gling, a computer vision engineer and Ph.D., Ph.D., from the Department of Automation at Tsinghua University, said that deep learning technology is based on a large number of examples, just like children collecting reality. The world's information is the same. Moreover, the more data "feed", the smarter it is and it won't "indigestion." Because big data is indispensable, the basics of deep learning at present are basically IT giants with large amounts of data, such as Google, Microsoft, Baidu, etc.
Nowadays, deep learning technology overcomes traditional machine learning methods in the fields of speech recognition, computer vision, language translation, etc., and even surpasses human recognition ability in face verification and image classification. Experts predict that in a few years, the phones in our pockets can also run a neural network as complex as the human brain.
However, as far as the current trend is concerned, deep learning technology is still not a substitute for “people sitting in the back-end monitoring roomâ€. For example, if you and your friends rush to check out after dining in a restaurant, this smart process is still difficult to judge whether it is fighting or something. It can be seen that logical judgment and emotional choice are obstacles that are difficult to overcome in deep learning.
The case can identify the bad guy's system at a glance
The technology company, which specializes in computer vision and artificial intelligence, is deeply involved in the application of intelligent learning systems based on deep learning technology to the field of banking security monitoring.
Considering that traditional optical lenses lose the “depth†dimension when recognizing images, Gling deeply developed a three-dimensional sensor specifically for bank security. Behind it, an algorithmic model trained by a reward and punishment mechanism can actively identify anomalies. "I saw someone approaching a someone's ATM, not the one next to it. At this time, to identify his trajectory and judge whether his behavior is normal, it involves deep learning." Ginger savvy CEO He Bofei introduced If the system recognizes an exception, it will be pushed to the background supervisor. In order to teach the machine to judge accurately, it is necessary to provide hundreds of thousands of pictures of the data behind.
He Bofei pointed out that giving the smart recognition system a side face or a full body photo without a face, it can also quickly lock the target with an accuracy of more than 99%. The premise is to build a sample library of 6,000 to 15,000. "Once the sample reaches a million levels, the accuracy may be reduced by 20% or more."
Banknote Value Counting Machine Two CIS
Suzhou Ribao Technology Co. Ltd. , https://www.ribaoeurope.com