g***e
发帖数: 577 | 1 train AI with data like:
[x1,x2,y] -> z where z = 1 if y = x1+x2, otherwise z = 0.
and your goal is to make the machine predict
[x1,x2,y] -> 0 or 1 ?
whats the best choice of method here. Condition is you can't write the logic
of y = x1+x2 explicitly in your model ( ie, you can't use the prior
knowledge of this, instead you would need the machine to discover this via
general algorithms ). | y********g
发帖数: 81 | 2 train z ~ x1, x2, (x1+x2-y) using a decision tree model. It will find two
splits for the third variable around 0. | g***e
发帖数: 577 | 3 thank you very much!
【在 y********g 的大作中提到】
: train z ~ x1, x2, (x1+x2-y) using a decision tree model. It will find two
: splits for the third variable around 0.
| l***e
发帖数: 108 | 4 我觉得你这第三个variable是利用了prior knowledge
【在 y********g 的大作中提到】
: train z ~ x1, x2, (x1+x2-y) using a decision tree model. It will find two
: splits for the third variable around 0.
| l****r
发帖数: 119 | | f*******6
发帖数: 56 | 6 [1 0
0 1
系数1 系数2] X [x1 X2]
当系数1，系数2同时为1时才能得出Z=1
但问题本身是多重共线性，是不是就无法求出系数1，系数2？ | v******y
发帖数: 4134 | | f*******6
发帖数: 56 | 8 我是新手，挺喜欢看到这样的帖子，大家讨论，如果高手不吝赐教，潜在的帮助更多的
中国学生学习。好事呀。
【在 v******y 的大作中提到】
: 为什么有人免费帮他做题啊？
|
|
