What is Artificial Intelligence? Popular terms, is to make the machine can think like humans. The need to explain too much, because we have a variety of science fiction movies of artificial intelligence is very familiar. Now everyone should be interested in - how artificial intelligence?
From the summer of 1956, first proposed "artificial intelligence" of the term began, scientists have tried various methods to achieve it. These methods include expert systems, decision trees, inductive logic, clustering, etc., but these are false intelligence. Until the emergence of artificial neural network technology, just let the machine have a "real smart."
Why are false intelligence methods before it? Because we humans can clearly know their internal analysis process, they are just a large, complex program only; and artificial neural network is different, it's inside a black box, just like our human brain, we do not know its internal analysis, we do not know how it is to identify a person's face, do not know how it go beat world champion. We're just as it constructed a shell of it, like human beings, we just gave birth to a child only, his mind is how to think we do not know! This is the AI's scary, because in the future it is likely that we humans should not live in this world, and our destroyed; to this end, the world has set up a number of security association to guard against artificial intelligence.
Artificial neural networks are inspired by the human brain structure and created, which is the root cause of it has really intelligent. In our brain, billions of cells called neurons, which are connected into a neural network.
Artificial neural network is the imitation of the above network structure. The following is a configuration diagram of an artificial neural network. Each circle represents a neuron, they are connected together to form a network.
Dendritic cells of human brain neurons receive different intensities from a plurality of external stimulation, and neuronal cells in vivo treatment, then converted to an output. As shown below.
Artificial neurons have similar works. As shown below.
The above x is an input neurons, corresponding to the plurality of received external stimulus dendrites. w is the weight of each weight corresponding to an input, which affects the stimulation intensity for each input x.
The simpler structure of the brain, so the lower the IQ. Single-celled organisms are the lowest IQ. Artificial neural networks are the same, the more complex the network, the more powerful it is, so we need depth neural network. Depth here refers to the number of layers, the more and more layers so the more complex structure of the neural network.
训练深度神经网络的过程就叫做深度学习。网络构建好了后，我们只需要负责不停地将训练数据输入到神经网络中，它内部就会自己不停地发生变化不停地学习。打比方说我们想要训练一个深度神经网络来识别猫。我们只需要不停地将猫的图片输入到神经网络中去。训练成功后，我们任意拿来一张新的图片，它都能判断出里面是否有猫。但我们并不知道他的分析过程是怎样的，它是如何判断里面是否有猫的。就像当我们教小孩子认识猫时，我们拿来一些白猫，告诉他这是猫，拿来一些黑猫，告诉他这也是猫，他脑子里会自己不停地学习猫的特征。最后我们拿来一些花猫，问他，他会告诉你这也是猫。但他是怎么知道的？他脑子里的分析过程是怎么样的？我们无从知道~~
通过对本篇文章的学习，我们知道了通过人工神经网络可以实现真正的人工智能。下一小节我就详细地为大家讲解神经网络。