机器学习中使用的神经网络(二)

什么是神经网络？

大脑如何工作？

Each neuron receives inputs from other neurons  每个神经元接收其他神经元的输入
- A few neurons also connect to receptors.  一些神经元连接到神经末梢
- Cortical neurons use spikes to communicate.大脑神经元使用突触来通讯
• The effect of each input line on the neuron is controlled by a synaptic weight

    每行输入神经元的效果都由突触的权值控制

– The weights can be positive or negative. 权值可能是正的也可以是负的
• The synaptic weights adapt so that the whole network learns to perform useful computations

　　突触的权值调整，是整个网络进行有用的计算

– Recognizing objects, understanding language, making plans,controlling the body.

    识别物体，理解语言，规划，控制身体

– A huge number of weights can affect the computation in a very short time. Much better bandwidth to stored knowlege than a modern workstation has.
  大量的权值短时间内能够影响计算。

Modularity and the brain 模块性大脑
• Different bits of the cortex do different things. 皮层的不同部位做不同的事
– Local damage to the brain has specific effects.　　局部的损伤导致某些特定的影响，比如失去语言能力，不能识别
– Specific tasks increase the blood flow to specific regions.
• But the remarkable thing about cortex is looks pretty much the same all over, and that strongly suggests that it's got a fairly flexible universal learning alogrithm on it.
– Early brain damage makes functions relocate.
• Cortex is made of general purpose stuff that has the ability to turn into
special purpose hardware in response to experience.
– This gives rapid parallel computation plus flexibility.
– Conventional computers get flexibility by having stored sequential programs, but this requires very fast central processors to perform long sequential computations.