谁能给较为完整科普下water-filling strategy阿，谢谢。

这id。。。

就是一个约束优化问题而已，可以看做一个点到单纯形上的射影，推荐你看这篇论文“Asynchronous Iterative Water-Filling for Gaussian Frequency-Selective Interference Channels”，一下就豁然开朗了。

From the optimization point of view, water-filling is a good demonstration o
f the usefulness of KKT conditions. The optimization problem in consideratio
n aims at optimizing the total capacity of n channels, each with a different
quality (represented by a distinct noise power), under the constraint that
the total transmit power is limited. Therefore, mathematically, the objectiv
e of the problem is a sum-log function (as capacity per channel is given by
the Shannon formula and the total capacity is the sume of individual channel
capacities) and the contraint is such that the sum of all powers (one per c
hannel) is bounded by a constant. If you write the KKT conditions of this pr
oblem, you will find a "threshold". For a given channel, if the noise power
is beyond this threshold, the channel is not used (power allocation is zero)
; otherwise the allocated power to a given channel is the difference between
the threshold and the noise power of the channel.

这只是water-filling在并行信道上的应用例子，其实这种对数求和的优化很多其它地方也用得到，呵呵

Well, I assume that is what LZ was asking. If you go to a different research
area, water-filling might mean very different things. For example, a common
algorithm to solve max-min fairness in network resource allocation is calle
d water-filling, but it really has nothing to do with what people talk about
in communcation.

re
他这个描述方式的确很有意思啊

你的ID和你提的问题同样令人震撼!