10 - 1 - Deciding What to Try Next (6 min).srt

1
00:00:00,300 --> 00:00:02,290
By now you have seen a lot of different learning algorithms.
到目前为止 我们已经介绍了许多不同的学习算法
(字幕整理：中国海洋大学 黄海广,haiguang2000@qq.com )

2
00:00:03,330 --> 00:00:04,450
And if you've been following along
如果你一直跟着这些视频的进度学习

3
00:00:04,770 --> 00:00:06,030
these videos you should consider
你会发现自己已经不知不觉地

4
00:00:06,770 --> 00:00:09,530
yourself an expert on many state-of-the-art machine learning techniques.
成为一个了解许多先进机器学习技术的专家了

5
00:00:09,730 --> 00:00:12,310
But even among
然而 在懂机器学习的人当中

6
00:00:12,560 --> 00:00:14,460
people that know a certain learning algorithm.
依然存在着很大的差距

7
00:00:15,250 --> 00:00:16,830
There's often a huge difference between
一部分人确实掌握了

8
00:00:17,090 --> 00:00:18,240
someone that really knows how
怎样高效有力地

9
00:00:18,410 --> 00:00:20,130
to powerfully and effectively apply
运用这些学习算法

10
00:00:20,450 --> 00:00:22,270
that algorithm, versus someone that's
而另一些人

11
00:00:22,950 --> 00:00:24,090
less familiar with some of
他们可能对我马上要讲的东西

12
00:00:24,160 --> 00:00:25,080
the material that I'm about
就不是那么熟悉了

13
00:00:25,420 --> 00:00:26,900
to teach and who doesn't really
他们可能没有完全理解

14
00:00:27,080 --> 00:00:28,090
understand how to apply these
怎样运用这些算法

15
00:00:28,250 --> 00:00:29,180
algorithms and can end up
因此总是

16
00:00:29,570 --> 00:00:30,760
wasting a lot of
把时间浪费在

17
00:00:30,870 --> 00:00:33,320
their time trying things out that don't really make sense.
毫无意义的尝试上

18
00:00:34,380 --> 00:00:35,180
What I would like to do is
我想做的是

19
00:00:35,340 --> 00:00:36,350
make sure that if you
确保你在设计

20
00:00:36,560 --> 00:00:37,830
are developing machine learning systems,
机器学习的系统时

21
00:00:38,600 --> 00:00:39,780
that you know how to choose
你能够明白怎样选择

22
00:00:40,400 --> 00:00:42,900
one of the most promising avenues to spend your time pursuing.
一条最合适 最正确的道路

23
00:00:43,890 --> 00:00:45,050
And on this and the next
因此 在这节视频

24
00:00:45,190 --> 00:00:46,530
few videos I'm going to
和之后的几段视频中

25
00:00:46,750 --> 00:00:47,890
give a number of practical
我将向你介绍一些实用的

26
00:00:48,380 --> 00:00:51,150
suggestions, advice, guidelines on how to do that.
建议和指导 帮助你明白怎样进行选择

27
00:00:51,520 --> 00:00:53,410
And concretely what we'd
具体来讲

28
00:00:53,600 --> 00:00:54,460
focus on is the problem
我将重点关注的问题是

29
00:00:54,940 --> 00:00:56,380
of, suppose you are
假如你在开发

30
00:00:56,580 --> 00:00:57,760
developing a machine learning system
一个机器学习系统

31
00:00:58,390 --> 00:00:59,390
or trying to improve the performance
或者想试着改进

32
00:00:59,950 --> 00:01:01,810
of a machine learning system, how
一个机器学习系统的性能

33
00:01:02,000 --> 00:01:03,630
do you go about deciding what are
你应如何决定

34
00:01:03,700 --> 00:01:05,260
the proxy avenues to try
接下来应该

35
00:01:07,620 --> 00:01:07,620
next?
选择哪条道路？

36
00:01:09,300 --> 00:01:11,200
To explain this, let's continue using
为了解释这一问题

37
00:01:11,670 --> 00:01:13,210
our example of learning to
我想仍然使用

38
00:01:13,350 --> 00:01:15,280
predict housing prices.
预测房价的学习例子

39
00:01:15,570 --> 00:01:17,760
And let's say you've implement and regularize linear regression.
假如你已经完成了正则化线性回归

40
00:01:18,700 --> 00:01:20,090
Thus minimizing that cost function
也就是最小化代价函数J的值

41
00:01:20,520 --> 00:01:22,870
j.  Now suppose that
假如

42
00:01:23,130 --> 00:01:24,310
after you take your learn parameters,
在你得到你的学习参数以后

43
00:01:24,820 --> 00:01:26,570
if you test your hypothesis on
如果你要将你的假设函数

44
00:01:26,720 --> 00:01:28,360
the new set of houses, suppose you
放到一组新的房屋样本上进行测试

45
00:01:28,540 --> 00:01:29,470
find that this is making huge
假如说你发现在预测房价时

46
00:01:29,860 --> 00:01:31,770
errors in this prediction of the housing prices.
产生了巨大的误差

47
00:01:33,220 --> 00:01:34,490
The question is what should
现在你的问题是 要想改进这个算法

48
00:01:34,670 --> 00:01:37,600
you then try mixing in order to improve the learning algorithm?
接下来应该怎么办？

49
00:01:39,000 --> 00:01:40,000
There are many things that one
实际上你可以想出

50
00:01:40,210 --> 00:01:41,460
can think of that could improve
很多种方法来改进

51
00:01:41,950 --> 00:01:43,660
the performance of the learning algorithm.
这个算法的性能

52
00:01:44,800 --> 00:01:47,510
One thing they could try, is to get more training examples.
其中一种办法是 使用更多的训练样本

53
00:01:48,060 --> 00:01:49,240
And concretely, you can imagine, maybe, you
具体来讲 也许你能想到

54
00:01:49,600 --> 00:01:51,150
know, setting up phone surveys, going
通过电话调查

55
00:01:51,570 --> 00:01:52,820
door to door, to try to
或上门调查

56
00:01:52,930 --> 00:01:54,050
get more data on how much
来获取更多的

57
00:01:54,310 --> 00:01:56,660
different houses sell for.
不同的房屋出售数据

58
00:01:57,730 --> 00:01:58,770
And the sad thing is I've seen
遗憾的是

59
00:01:59,010 --> 00:02:00,060
a lot of people spend a
我看到好多人花费了好多时间

60
00:02:00,200 --> 00:02:01,400
lot of time collecting more training
想收集更多的训练样本

61
00:02:01,760 --> 00:02:03,270
examples, thinking oh, if we have
他们总认为 噢 要是我有

62
00:02:03,760 --> 00:02:04,780
twice as much or ten times
两倍甚至十倍数量的训练数据

63
00:02:05,050 --> 00:02:07,250
as much training data, that is certainly going to help, right?
那就一定会解决问题的 是吧？

64
00:02:07,990 --> 00:02:09,020
But sometimes getting more training
但有时候 获得更多的训练数据

65
00:02:09,380 --> 00:02:10,680
data doesn't actually help
实际上并没有作用

66
00:02:11,240 --> 00:02:11,920
and in the next few videos
在接下来的几段视频中

67
00:02:12,430 --> 00:02:13,670
we will see why, and we
我们将解释原因

68
00:02:13,720 --> 00:02:15,270
will see how you
我们也将知道

69
00:02:15,500 --> 00:02:16,780
can avoid spending a lot
怎样避免把过多的时间

70
00:02:16,950 --> 00:02:18,160
of time collecting more training data
浪费在收集更多的训练数据上

71
00:02:18,910 --> 00:02:20,660
in settings where it is just not going to help.
这实际上是于事无补的

72
00:02:22,370 --> 00:02:23,510
Other things you might try are
另一个方法 你也许能想到的

73
00:02:23,730 --> 00:02:25,830
to well maybe try a smaller set of features.
是尝试选用更少的特征集

74
00:02:26,470 --> 00:02:27,270
So if you have some set
因此如果你有一系列特征

75
00:02:27,450 --> 00:02:29,030
of features such as x1,
比如x1

76
00:02:29,270 --> 00:02:30,330
x2, x3 and so on,
x2 x3等等

77
00:02:30,680 --> 00:02:31,840
maybe a large number of features.
也许有很多特征

78
00:02:32,570 --> 00:02:33,460
Maybe you want to spend time
也许你可以花一点时间

79
00:02:33,860 --> 00:02:35,240
carefully selecting some small
从这些特征中

80
00:02:35,590 --> 00:02:37,410
subset of them to prevent overfitting.
仔细挑选一小部分来防止过拟合

81
00:02:38,670 --> 00:02:40,730
Or maybe you need to get additional features.
或者也许你需要用更多的特征

82
00:02:41,330 --> 00:02:42,390
Maybe the current set of features
也许目前的特征集

83
00:02:42,570 --> 00:02:44,740
aren't informative enough and you
对你来讲并不是很有帮助

84
00:02:44,840 --> 00:02:47,460
want to collect more data in the sense of getting more features.
你希望从获取更多特征的角度 来收集更多的数据

85
00:02:48,510 --> 00:02:49,590
And once again this is the
同样地

86
00:02:49,730 --> 00:02:50,900
sort of project that can scale
你可以把这个问题

87
00:02:51,180 --> 00:02:52,260
up the huge projects can you
扩展为一个很大的项目

88
00:02:52,580 --> 00:02:54,110
imagine getting phone
比如使用电话调查

89
00:02:54,350 --> 00:02:55,280
surveys to find out more
来得到更多的房屋案例

90
00:02:55,490 --> 00:02:57,230
houses, or extra land
或者再进行土地测量

91
00:02:57,640 --> 00:02:58,620
surveys to find out more
来获得更多有关

92
00:02:58,800 --> 00:03:01,130
about the pieces of land and so on, so a huge project.
这块土地的信息等等 因此这是一个复杂的问题

93
00:03:01,690 --> 00:03:02,820
And once again it would be
同样的道理

94
00:03:02,930 --> 00:03:04,140
nice to know in advance if
我们非常希望

95
00:03:04,330 --> 00:03:05,210
this is going to help before we
在花费大量时间完成这些工作之前

96
00:03:05,760 --> 00:03:07,690
spend a lot of time doing something like this.
我们就能知道其效果如何

97
00:03:07,920 --> 00:03:09,390
We can also try
我们也可以尝试

98
00:03:10,360 --> 00:03:12,100
adding polynomial features things like
增加多项式特征的方法

99
00:03:12,180 --> 00:03:13,100
x2 square x2 square and product
比如x1的平方 x2的平方

100
00:03:13,860 --> 00:03:14,700
features x1, x2.
x1 x2的乘积

101
00:03:14,930 --> 00:03:16,040
We can still spend quite a
我们可以花很多时间

102
00:03:16,180 --> 00:03:17,830
lot of time thinking about that and
来考虑这一方法

103
00:03:18,270 --> 00:03:19,340
we can also try other things like
我们也可以考虑其他方法

104
00:03:19,540 --> 00:03:21,390
decreasing lambda, the regularization parameter or increasing lambda.
减小或增大正则化参数lambda的值

105
00:03:23,840 --> 00:03:25,160
Given a menu of options
我们列出的这个单子

106
00:03:25,520 --> 00:03:26,680
like these, some of which
上面的很多方法

107
00:03:26,970 --> 00:03:28,240
can easily scale up to
都可以扩展开来

108
00:03:28,950 --> 00:03:30,000
six month or longer projects.
扩展成一个六个月或更长时间的项目

109
00:03:31,310 --> 00:03:32,660
Unfortunately, the most common
遗憾的是

110
00:03:32,760 --> 00:03:34,010
method that people use to
大多数人用来选择这些方法的标准

111
00:03:34,170 --> 00:03:36,040
pick one of these is to go by gut feeling.
是凭感觉

112
00:03:36,520 --> 00:03:37,670
In which what many people
也就是说

113
00:03:38,170 --> 00:03:39,520
will do is sort of randomly
大多数人的选择方法是

114
00:03:39,940 --> 00:03:41,100
pick one of these options and
随便从这些方法中选择一种

115
00:03:41,250 --> 00:03:43,050
maybe say, "Oh, lets go and get more training data."
比如他们会说 “噢 我们来多找点数据吧”

116
00:03:43,980 --> 00:03:45,480
And easily spend six months collecting
然后花上六个月的时间

117
00:03:45,880 --> 00:03:47,540
more training data or maybe someone
收集了一大堆数据

118
00:03:47,780 --> 00:03:48,860
else would rather be saying, "Well,
然后也许另一个人说

119
00:03:49,430 --> 00:03:51,810
let's go collect a lot more features on these houses in our data set."
“好吧 让我们来从这些房子的数据中多找点特征吧”

120
00:03:52,780 --> 00:03:54,010
And I have a lot
我很遗憾不止一次地看到

121
00:03:54,220 --> 00:03:55,870
of times, sadly seen people spend, you know,
很多人花了

122
00:03:56,630 --> 00:03:58,360
literally 6 months doing one
不夸张地说 至少六个月时间

123
00:03:58,530 --> 00:03:59,680
of these avenues that they have
来完成他们随便选择的一种方法

124
00:04:00,240 --> 00:04:01,810
sort of at random only to
而在六个月或者更长时间后

125
00:04:01,920 --> 00:04:03,220
discover six months later that
他们很遗憾地发现

126
00:04:03,460 --> 00:04:05,610
that really wasn't a promising avenue to pursue.
自己选择的是一条不归路

127
00:04:07,090 --> 00:04:08,170
Fortunately, there is a
幸运的是

128
00:04:08,310 --> 00:04:10,650
pretty simple technique that can
有一系列简单的方法

129
00:04:10,930 --> 00:04:12,640
let you very quickly rule
能让你事半功倍

130
00:04:12,900 --> 00:04:14,190
out half of the things
排除掉单子上的

131
00:04:14,500 --> 00:04:16,440
on this list as being potentially
至少一半的方法

132
00:04:16,970 --> 00:04:17,990
promising things to pursue.
留下那些确实有前途的方法

133
00:04:18,390 --> 00:04:19,310
And there is a very simple technique,
同时也有一种很简单的方法

134
00:04:19,830 --> 00:04:21,080
that if you run, can easily
只要你使用

135
00:04:21,710 --> 00:04:22,820
rule out many of these options,
就能很轻松地排除掉很多选择

136
00:04:24,120 --> 00:04:25,470
and potentially save you
从而为你节省

137
00:04:25,580 --> 00:04:28,600
a lot of time pursuing something that's just is not going to work.
大量不必要花费的时间

138
00:04:29,610 --> 00:04:30,950
In the next two videos
在接下来的两段视频中

139
00:04:31,320 --> 00:04:32,450
after this, I'm going to
我首先介绍

140
00:04:32,560 --> 00:04:35,420
first talk about how to evaluate learning algorithms.
怎样评估机器学习算法的性能

141
00:04:36,540 --> 00:04:37,810
And in the next few
然后在之后的几段视频中

142
00:04:38,080 --> 00:04:39,770
videos after that, I'm
我将开始

143
00:04:40,070 --> 00:04:41,130
going to talk about these techniques,
讨论这些方法

144
00:04:42,470 --> 00:04:44,270
which are called the machine learning diagnostics.
它们也被称为"机器学习诊断法"

145
00:04:46,690 --> 00:04:47,980
And what a diagnostic is, is
“诊断法”的意思是

146
00:04:48,120 --> 00:04:49,080
a test you can run,
这是一种测试法

147
00:04:49,900 --> 00:04:52,240
to get insight into what
你通过执行这种测试

148
00:04:52,430 --> 00:04:53,740
is or isn't working with
能够深入了解

149
00:04:54,130 --> 00:04:55,810
an algorithm, and which will
某种算法到底是否有用

150
00:04:56,070 --> 00:04:57,720
often give you insight as to
这通常也能够告诉你

151
00:04:57,940 --> 00:04:59,360
what are promising things to try
要想改进一种算法的效果

152
00:04:59,920 --> 00:05:01,100
to improve a learning algorithm's
什么样的尝试

153
00:05:03,910 --> 00:05:03,910
performance.
才是有意义的

154
00:05:04,730 --> 00:05:07,140
We'll talk about specific diagnostics later in this video sequence.
在这一系列的视频中我们将介绍具体的诊断法

155
00:05:08,050 --> 00:05:09,230
But I should mention in advance
但我要提前说明一点的是

156
00:05:09,440 --> 00:05:10,780
that diagnostics can take
这些诊断法的执行和实现

157
00:05:11,100 --> 00:05:12,280
time to implement and can sometimes,
是需要花些时间的

158
00:05:12,820 --> 00:05:14,300
you know, take quite a
有时候

159
00:05:14,340 --> 00:05:15,610
lot of time to implement and
确实需要花很多时间

160
00:05:15,740 --> 00:05:17,120
understand but doing so
来理解和实现

161
00:05:17,410 --> 00:05:18,330
can be a very good use
但这样做的确是

162
00:05:18,610 --> 00:05:19,380
of your time when you are
把时间用在了刀刃上

163
00:05:19,660 --> 00:05:21,460
developing learning algorithms because they
因为这些方法

164
00:05:21,560 --> 00:05:22,660
can often save you from
让你在开发学习算法时

165
00:05:22,880 --> 00:05:24,670
spending many months pursuing an
节省了几个月的时间

166
00:05:24,840 --> 00:05:26,580
avenue that you could
早点从不必要的尝试中解脱出来

167
00:05:26,870 --> 00:05:29,460
have found out much earlier just was not going to be fruitful.
早日脱离苦海

168
00:05:32,220 --> 00:05:33,070
So in the next few
因此

169
00:05:33,250 --> 00:05:34,250
videos, I'm going to first
在接下来几节课中

170
00:05:34,570 --> 00:05:36,220
talk about how evaluate your
我将先来介绍

171
00:05:36,450 --> 00:05:38,210
learning algorithms and after
如何评价你的学习算法

172
00:05:38,410 --> 00:05:39,210
that I'm going to talk
在此之后

173
00:05:39,300 --> 00:05:41,490
about some of these diagnostics which will hopefully
我将介绍一些诊断法

174
00:05:41,810 --> 00:05:42,950
let you much more
希望能让你更清楚

175
00:05:43,110 --> 00:05:44,470
effectively select more of the
在接下来的尝试中

176
00:05:44,770 --> 00:05:45,880
useful things to try mixing
如何选择更有意义的方法

177
00:05:46,560 --> 00:05:48,200
if your goal to improve
最终达到改进机器学习系统性能的目的

178
00:05:48,760 --> 00:05:50,430
the machine learning system.