t**r 发帖数: 3428 | 1 Nothing is new. Everything is in pattern recognition area.
And ML is not that useful.
Better to take time to study distributed system . |
k**********g 发帖数: 989 | 2
Not new > not useful.
【在 t**r 的大作中提到】 : Nothing is new. Everything is in pattern recognition area. : And ML is not that useful. : Better to take time to study distributed system .
|
c****e 发帖数: 1453 | 3 Similar to cloud, it's not the technology that is fundamentally new, but the
impact on application/business is new.
Similar to big data, it's not about it really boosts your sales or not, it's
the mentality that you believe it will, and you even change your business
process to incorporate. |
Y**G 发帖数: 1089 | |
N******K 发帖数: 10202 | 5 还是有用 就是吹得过火 尤其是什么 learn 这个词 根本就是挂羊头卖狗肉
【在 t**r 的大作中提到】 : Nothing is new. Everything is in pattern recognition area. : And ML is not that useful. : Better to take time to study distributed system .
|
Y**G 发帖数: 1089 | 6 难道FB里面拿30万年薪的Data Scientist不做ML难道是混吃的?
【在 N******K 的大作中提到】 : 还是有用 就是吹得过火 尤其是什么 learn 这个词 根本就是挂羊头卖狗肉
|
N******K 发帖数: 10202 | 7 现在的算法 根本不能称为“学习”
【在 Y**G 的大作中提到】 : 难道FB里面拿30万年薪的Data Scientist不做ML难道是混吃的?
|
m********5 发帖数: 17667 | 8 是混吃的
远不如招10个学统计的有用
【在 Y**G 的大作中提到】 : 难道FB里面拿30万年薪的Data Scientist不做ML难道是混吃的?
|
c*******9 发帖数: 9032 | 9 Data Scientist 很多是学统计吧,不过有的银行更喜欢纯数学的。一般学统计的学的
太浅。
【在 m********5 的大作中提到】 : 是混吃的 : 远不如招10个学统计的有用
|
m********5 发帖数: 17667 | 10 招数学的肯定更有用
不过那种屌丝IT公司用得来么?
data scientist少数是精通统计的
大多是半路出家两头不行
【在 c*******9 的大作中提到】 : Data Scientist 很多是学统计吧,不过有的银行更喜欢纯数学的。一般学统计的学的 : 太浅。
|
|
|
T*****u 发帖数: 7103 | 11 公司里面用的都是practical machine learning,一部分原因是theoretical machine
learning does not work on real data,还有一部分原因是real statisticians dont
know how to program |
c********l 发帖数: 8138 | 12 re这个
machine
dont
【在 T*****u 的大作中提到】 : 公司里面用的都是practical machine learning,一部分原因是theoretical machine : learning does not work on real data,还有一部分原因是real statisticians dont : know how to program
|
w***g 发帖数: 5958 | 13 不知道machine learning还有这个分类。不知道SVM和naive bayesian属于practical还是
theoretical。还是说不干活的就是theoretical?
machine
dont
【在 T*****u 的大作中提到】 : 公司里面用的都是practical machine learning,一部分原因是theoretical machine : learning does not work on real data,还有一部分原因是real statisticians dont : know how to program
|
N******K 发帖数: 10202 | 14 什么叫精通统计
统计这个东西 门槛很低 数学中的文科
【在 m********5 的大作中提到】 : 是混吃的 : 远不如招10个学统计的有用
|
k**********g 发帖数: 989 | 15
还是
有 Supervised ML, Unsupervised ML 之分。
实际应用中,也有 Artificial Intelligence 而不属於 Machine Learning 的例子。
全手工建模 (Mathematical Modeling) 就是 「Human Machine Learning」, the data
analyst (mathematical modeler) doing 100% of the job of making the model.
Today's (renewed efforts) on Machine Learning should be narrowly defined as
Deep Learning, where no examples are given to the algorithm, and the
algorithm figures 100% of the model by itself.
This is what makes Andrew Ng's recent experiments ground-breaking.
【在 w***g 的大作中提到】 : 不知道machine learning还有这个分类。不知道SVM和naive bayesian属于practical还是 : theoretical。还是说不干活的就是theoretical? : : machine : dont
|
N******K 发帖数: 10202 | 16 加拿大 hilton的deep learning
yanlelun
这个ng 也就是个随大流的
data
as
【在 k**********g 的大作中提到】 : : 还是 : 有 Supervised ML, Unsupervised ML 之分。 : 实际应用中,也有 Artificial Intelligence 而不属於 Machine Learning 的例子。 : 全手工建模 (Mathematical Modeling) 就是 「Human Machine Learning」, the data : analyst (mathematical modeler) doing 100% of the job of making the model. : Today's (renewed efforts) on Machine Learning should be narrowly defined as : Deep Learning, where no examples are given to the algorithm, and the : algorithm figures 100% of the model by itself. : This is what makes Andrew Ng's recent experiments ground-breaking.
|
V*********r 发帖数: 666 | 17 good one
【在 Y**G 的大作中提到】 : 难道FB里面拿30万年薪的Data Scientist不做ML难道是混吃的?
|
n******t 发帖数: 4406 | 18 come on, you know this is bullshit.
the
's
【在 c****e 的大作中提到】 : Similar to cloud, it's not the technology that is fundamentally new, but the : impact on application/business is new. : Similar to big data, it's not about it really boosts your sales or not, it's : the mentality that you believe it will, and you even change your business : process to incorporate.
|
m****i 发帖数: 3406 | 19 这个所谓“algorithm figures 100% of the model by itself.”也要通过
一定的预先定好模式的吧,只是模式抽象层次的问题。
相对于搞ML的人数,每年巨大的paper数量,ML这个领域对现实
的impact应该是很小的。
data
as
【在 k**********g 的大作中提到】 : : 还是 : 有 Supervised ML, Unsupervised ML 之分。 : 实际应用中,也有 Artificial Intelligence 而不属於 Machine Learning 的例子。 : 全手工建模 (Mathematical Modeling) 就是 「Human Machine Learning」, the data : analyst (mathematical modeler) doing 100% of the job of making the model. : Today's (renewed efforts) on Machine Learning should be narrowly defined as : Deep Learning, where no examples are given to the algorithm, and the : algorithm figures 100% of the model by itself. : This is what makes Andrew Ng's recent experiments ground-breaking.
|