We consider kernel desnity estimation using the stagewise minimization algorithm with its dictionary having various bandwidths. Let X_i~T =(X_(i1),X_(i2),…,X_(id)), i =1,2,...,m + n, be d-dimensional i.i.d. sample generated from the d-dimensional density finction f. For this i.i.d. sample, we define X*_i≡X_i =1,2,...,m, and use them for the dictionary. The rest of the sample, X_i, i = m +1,m + 2,...,m + n, are utilized for the algorithm of density estimation. Denoting B to be a set of d x dimensional bandwidth matrices H_j, j =1,2,...,|B|, we define the dictionary, D={ΦH_j(·-X*_i)|H_j ∈ B,i = 1,2,...,m,j =1,2,..., |B|},|D| = m ×|B|, where ΦH_j(·-X*_j) is a density function with its mean X*_j and variance-covariance matrix H_j. Each word in D is denoted to be Φ_s(x|X*),s =1,2,...,m ×|B|, where each index number s corresponds to a combination of X*_i and H_j one-to-one.
展开▼
机译:我们考虑使用具有各种带宽的分支最小化算法考虑内核Desnity估计。让x_i〜t =(x_(i1),x_(i2),...,x_(id)),i = 1,2,...,m + n,为d维i.i.d.d.从D尺寸密度的样品产生的样品。为此,i.D.样本,我们定义x *_i≡x_i= 1,2,...,m,并为字典使用它们。样品的其余部分X_I,I = M + 1,M + 2,...,M + N用于密度估计算法。表示B为一组DX尺寸带宽矩阵H_J,J = 1,2,...,| B |,我们定义字典,D = {φH_J(·-X * _I)| H_J∈B,i = 1,2,...,m,j = 1,2,...,| b |},| d | = m×| b |,其中φh_j(·-x * _j)是具有其平均x * _j和方差协方差矩阵h_j的密度函数。 d中的每个单词都表示为φ_s(x | x *),s = 1,2,...,m×| b |,其中每个索引号s对应于x * _i和h_j的组合一体化-一。
展开▼