例如,如果我们正在开发一个自动过滤系统来删除垃圾邮件,那么垃圾邮件过滤器就是我们所讨论的模型。
For instance, if we are developing anauto-filtering system to
remove spam mail, the spam mail filter is the modelthat we are
talking about.
在这个意义上,我们可以说模型是我们实际需要使用的。
In this sense, we can say the model is whatwe actually use.
有些人称模型为假设。
Some call the model a hypothesis.
这种称呼对于那些具有统计背景的人似乎更直观。
This term seems more intuitive to thosewith statistical
backgrounds.
机器学习并不是唯一的建模技术。
Machine Learning is not the only modelingtechnique.
在动力学领域,人们长期以来一直使用某种建模技术,它采用牛顿定律,把物体的运动描述为一系列称为运动方程的表达式。
In the field of dynamics, people have beenusing a certain modeling
technique, which employs Newton’s laws and describes the motion
ofobjects as a series of equations called equations of motion, for
a long time.
在人工智能领域,专家系统是基于专家知识和专业技术的问题求解模型。
In the field of Artificial Intelligence, wehave the expert system,
which is a problem-solving model that is based on theknowledge and
know-how of the experts.
这种模型如同专家一样工作。
The model works as well as the expertsthemselves.
然而,在某些领域,法律和逻辑推理对于建模不太有用。
However, there are some areas where lawsand logical reasoning are
not very useful for modeling.
涉及智能的典型问题,例如图像识别、语音识别和自然语言处理。
Typical problems can be found whereintelligence is involved, such
as image recognition, speech recognition, andnatural language
processing.
让我举个例子予以说明。
Let me give you an example.
请看图1-3并识别图中的数字。
Look at Figure 1-3 and identify thenumbers.
图1-3 当计算机没有识别模式时,它是如何辨识这些数字的呢?How
does acomputer identify numbers when they have no recognizable
pattern?
我相信你很快就能完成任务。
I’m sure you have completed the task in notime.
大多数人都能够顺利完成。
Most people do.
现在,我们让计算机来完成同样的事情。
Now, let’s make a computer do the samething.
我们该怎么办呢?
What do we do?
如果我们使用传统的建模技术,我们将需要找到一些规则或算法来区分这些手写的数字。
If we use a traditional modeling technique,we will need to find
some rule or algorithm to distinguish the written numbers.
嗯,我们为什么不应用你大脑中刚刚用过的规则来识别数字呢?
Hmm, why don’t we apply the rules that youhave just used to
identify the numbers in your brain?
应该很简单的,不是吗?
Easy enough, isn’t it?
然而,事实并非如此。
Well, not really.
这是一个非常具有挑战性的问题。
In fact, this is a very challengingproblem.
曾经,研究人员认为计算机要做到这一点只是小菜一碟,因为计算机具有比人类更快的运算速度。
There was a time when researchers thoughtit must be a piece of cake
for computers to do this, as it is very easy foreven a human and
computers are able to calculate much faster than humans.
——本文译自Phil Kim所著的《Matlab Deep Learning》
更多精彩文章请关注微信号: