Artificial intelligence has now penetrated into every aspect of life. From the attention-grabbing Go, chess opponents, to face recognition in mobile payment and mobile phone unlocking, artificial intelligence is used.
Artificial intelligence has many human perception and recognition capabilities, and can automatically perform certain tasks, such as recognizing sound, recognizing faces and making decisions.
How to automate tasks? Taking image recognition as an example, in the past, we entered a series of rules for the computer to make it rule-based (reale-based), which uses feature point categorization to identify images, but it is difficult to accurately identify.
If we tell the computer that there is a turnip on the left, it might think that the cone on the right is also a carrot.
To achieve true artificial intelligence, people have developed machine learning.
Machine learning is simply to write a set of algorithms that enable a computer to learn "automatically". We don't need to enter a lot of rules, just build a model with a small number of rules, and the computer can learn to create new data to analyze and judge new things.
By learning the 1 on the left, you can judge that the right side is also 1
We have previously demonstrated how to use a neural network to write a machine learning algorithm to automatically identify unknown images.
A 4-layer neural network input layer is output layer through several layers of algorithms
There are many ways to implement machine learning. The most widely discussed method in recent years is deep learning.
Deep learning is a technology for machine learning. It uses deep neural networks to implement machine learning. The depth of the network is tens or even hundreds of times deeper than the original network.
Enhanced learning can also achieve the effect of machine learning. Interested partners can search online, and there are no recommended books here.
a 34-layer deep neural network
What problems can this kind of network solve? The most popular of these is image recognition.
For example, after the computer gets some photos of cats, it can identify Chinese garden cats and other kinds of cats, and then classify them. This seemingly a wasteful use, if applied to the medical field, such as the resolution of good and diseased organs, or the current hot face recognition, will change human life.
Since 2010, in order to better develop image recognition technology, people have established ImageNet ImageNet, and even organized a database-based recognition contest ILSVRC, which is better than the image recognition method.
ImageNet database
Deep learning has achieved good results in the competition, so it has received more attention and development. Prior to this, the development was limited mainly because the chip and the graphics card did not move (the amount of calculation is too large).
Neodymium 6.5 Inch,Mid-Range Speaker,Pro Audio Midrange,Mid Range Pro Audio Speaker
Guangzhou BMY Electronic Limited company , https://www.bmy-speakers.com