基于密度峰值的聚類算法

1、Clustering by fast search and find of density peaks基于密度峰值的聚類算法 摘自2014-6-27Science期刊Clustering by fast search and find of density peaks算法思想算法思想Clustering by fast search and find of density peaks算法介紹算法介紹Clustering by fast search and find of density peaks算法介紹算法介紹Clustering by fast search and find of de

2、nsity peaks算法介紹算法介紹Clustering by fast search and find of density peaks算法介紹算法介紹123Clustering by fast search and find of density peaks聚類中心選取聚類中心選取Clustering by fast search and find of density peaks聚類中心選取聚類中心選取Clustering by fast search and find of density peaks聚類過程聚類過程類簇中心找到后，剩余的每個點被歸屬到它的有更高密度的最近鄰所屬類簇。

3、類簇分配只需一步即可完成，不像其它算法要對目標(biāo)函數(shù)進行迭代優(yōu)化。Clustering by fast search and find of density peaks算法存在的問題算法存在的問題下圖中的圖B（決策圖）對確定聚類中心具有決定性作用，但是，對聚類中心的選取需要人為干預(yù)，而且其中包含了主觀因素，不同的人可能選擇不同的點作為聚類中心。決策圖中聚類中心難以確定的例子Clustering by fast search and find of density peaks算法優(yōu)化算法優(yōu)化Clustering by fast search and find of density peaks算法存

4、在的問題算法存在的問題按照原文的聚類方法，一些分散的“離群點”也會被“強制”分到類簇當(dāng)中，造成聚類后類簇邊界不清晰，影響聚類效果。Clustering by fast search and find of density peaks算法優(yōu)化算法優(yōu)化Clustering by fast search and find of density peaks聚類結(jié)果展示聚類結(jié)果展示不區(qū)分cluster halo的聚類結(jié)果Clustering by fast search and find of density peaks聚類結(jié)果展示聚類結(jié)果展示區(qū)分cluster halo的聚類結(jié)果Clustering by fast search a