神经网路 空间映射 问题

樱风西日

* c: r6 i, `) m# V
$ A+ ?. O6 [2 E  G6 }. r

  g* N" }8 G: F% d# {7 w
% `$ }$ R, i& w) j9 `

% }9 A1 R' ~, o7 D. E* }2 m7 h; P
. S5 c7 X! T8 @6 @
* Z- O' \! K+ m: U( }

( e) ^# A' ^4 O0 H0 @7 q



! d: t+ p, T1 F% ]9 f2 X
# E* R3 V1 u" L: V
* f5 _: t% D/ R. v  J8 k+ u

) A+ h6 g. s0 N4 j/ u问题一：将神经网路训练完成之后 ，得到Sf与Sc之间的映射，即精细模型到粗模型之间的映射关系，即Xf到Xc的关系，即Xc=P（Xf），那么将粗模型最优值Xc*作为输入进行预测，为什么能得到下一次精细模型的解呢？

    我的理解应该是Xf=P-1(P的逆)（Xc*），类似于原始空间映射的方法。然而，难道这里是神经网络逆向预测吗？如果是，是怎么做的，能给些思路吗？

AuCu

如何学习神经网络
" Q2 T$ }! O+ S; P; M: e! @

lvjianchi

我得理解神经网络就是一种映射，有输入神经网络就会有输出，只不过得到的结果或许会偏差很大

小野兽

嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎刚嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎

小野兽

嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎刚嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎

小野兽

刚刚嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎刚嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎

小野兽

嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎刚嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎

小野兽

刚刚嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎刚嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎刚刚被

小野兽

嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎刚嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎

小野兽

嘎嘎嘎刚嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎刚嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎嘎