ADM-201 dump PMP dumps pdf SSCP exam materials CBAP exam sample questions

【人工智能实验系列】多维空间的可视化 – 译学馆
未登陆,请登陆后再发表信息
最新评论 (0)
播放视频

【人工智能实验系列】多维空间的可视化

A.I. Experiments: Visualizing High-Dimensional Space

[音乐]
[Music Play]
大家好 我是丹尼尔
Hi, I’m Daniel.
大家好 我是马丁
Hi, I’m Martin.
大家好 我是费尔南达
Hi, I’m Fernanda.
“数据视觉小分队”
“Small team of data visualizers”
计算机学习是十分困难的
Machine learning is pretty complex,
所以 我们一直在探索一种方法
So we’ve been experimenting with ways to
去想象发生了什么
visualize what’s happening.
[音乐]
[Music]
计算机学习中有一个核心观念
There’s a core concept in Machine Learning,
叫做:多思维空间
called high dimensional space.
有一种关于这个概念的方法让你头疼
Here’s one way to wrap your head around this concept.
你可以想想那些多思维的人
You can think about people as being high dimensional.
例如 女科学家
For example,take female scientists
你也可以想想
you can think about
他们的出生日期
when they were born
他们出生的地点
where they were born
他们如何学习的
their feels of the study
所有这些都属于人的多思维表现
each of these is like a dimension of that person
这时候 散发思维就会由困难变成混乱
These dimensions become difficult to tangle
当你考虑不同人的时候
when you think about different people.
因为有些人在某些方面可能是相似的
Because someone might be similar in someways
但又不同于他人
but very different in others
对于这样的一种情况
But this is a kind of thing
你就可以使用机械学习方法了
you can use Machine Learning for
用机器学习
With Machine Learning
电脑就会被告知这些思维的意思
the computer is told the meaning of these dimensions
就像数字那样
just sees them as numbers
它就会把每组数字看成数据点
And it sees each set of numbers as data point
但是仔细展望这些思维
but by looking across all of these dimensions of ones
在多思维空间可以把相关的数据点联系在一起
It’s able to play through related point closer together in high dimensional space
举一个具体的例子
Here is a concrete example
在一个单词里可以散发思维考虑
Where words treated in high dimensional data points
记住一件重要的事
The important thing to remember
我们没有告诉电脑单词的意思
is that we haven’t told the computer the meaning of words
可是 我们却看到了成千上万关于这个单词的例子
Instead,we have showing millions of sentences as a example so how words get used
有一个可视的例子
Here is a visualization of results
我们在寻找一个单词的子集
We are looking at a subset of words
电脑知道每点代表一个单词
the computer has learned about each dot represents one word
一个单词包含两百种思维的数据点
each word is a data point with two hundred the dimensions
使用一种叫降维的方法
using technique called T-SNE
电脑把那些相关的单词聚集在一起
the computer cluster words together that consider words releated
相似意思的单词也会被聚集在一起
and clusters form based meaning
即使我们从未讨论这些单词的意思
even though we never talk their the meaning of words
这有一组集合的数字
Here is a cluster of numbers
月份
months of the year
与空间相关的单词
words related to space
人名
people’s names
城市 等等
cities and so on
我们可以更紧密发现在一组单词中
we can also look closely at smaller set of words
如果我们查找钢琴
if we search “piano”
我们能使用降维工具帮我们寻找与钢琴相关的单词
we can run T-SNE only on words relate to piano
我们集合作曲流派
we get clusters of composers, genres
乐具
musical, instruments and more
此时 这种方法并不适用了
and this approach doesn’t just work for words
例如
For example
以多思维思考 你可以看做是一幅图案
you can also treat an image as a high-dimensional data point
这有一组数据
Here is a data set
许多人用各种方式从一写到九
a lot of people wrote digit between zero and nine ,people write all kinds of ways
所以 问题就是
so the question is
而不是随意操控密码规则
Instead of us needing to manually code the rule
对于人们写的方式
for all the ways people write
使用机器学习真的可以弄懂它本身嘛
could a machine figure it out itself use Machine Learning?
每个图像都对应时轴
each images seven hundreds data for pixels
电脑把每个像素看成一种思维
the computer treats each pixel as a dimension
然后 使用降维技术
again,using T-SNE
用多思维空间聚集这些图案
it cluster these images in high-dimensional space
我们按颜色区分好
we’ve color coded them
所以我们能更好地看见发生了什么
so it is easier for us to see what’s going on
我们可以看见成组的数据集合在一起
and you can see groups of digit clustering together
如果我们知道了这些数据的意思
if learn something about the meaning about these digits
我们一直在探寻的可视化技术
This visualization techniques we have been exploring
可在很多方面应用
can be use for all kinds of things
那就是我们致力于开发这项技术的原因
That’s why we are working on open sourcing
所有这些都属于”谷歌张量流图”
all of these as spread of tensorflow
所以不论谁都可以应用这项技术去探索他们在研究的
So that anyone can use these tool to explore their data

发表评论

译制信息
视频概述

介绍了高维空间的可视化,什么时候可以使用以及使用的范围。。

听录译者

小静子

翻译译者

小静子

审核员

自动通过审核

视频来源

https://www.youtube.com/watch?v=wvsE8jm1GzE

相关推荐