Question

我正在Matlab中实现粒子过滤器。如果我尝试对代码进行矢量化处理，则会收到上述错误。但是，如果我不进行向量化，它将很好地工作。我需要对其向量化，以节省for循环所花费的时间。任何帮助，将不胜感激。我已经浏览过其他文章来解决相同的问题，但是无法获得有关我的眼周问题的任何有用信息。

矢量化代码

n = size(X_kmin1,1);

X_k = proc_f(X_kmin1) + mvnrnd(zeros(n,1),proc_Q)';
ykp =  meas_h(X_k);

pyx = exp(-0.5 * (yk-ykp)' * ((meas_R)^-1) * (yk-ykp));
W_k = W_kmin1*pyx;

Wt = sum(W_k);
W_k = W_k/Wt;

错误声明

Out of memory. Type "help memory" for your options. 
Error in pfFilterStep (line 24) 
pyx = exp(-0.5 * ((yk-ykp)' * (meas_R)^-1 * (yk-ykp))); 
Error in Test1 (line 32) 
[X_k, W_k] = pfFilterStep(X_kmin1, W_kmin1, yk, proc_f, proc_Q, meas_h, meas_R);

工作代码

%   Input:
%   X_kmin1     [n x N] Particles for state x in time k-1
%   W_kmin1     [1 x N] Weights for state x in time k-1
%   y_k         [m x 1] Measurement vector for time k
%   proc_f      Handle for process function f(x_k-1)
%   proc_Q      [n x n] process noise covariance
%   meas_h      Handle for measurement model function h(x_k)
%   meas_R      [m x m] measurement noise covariance
%
% Output:
%   X_k         [n x N] Particles for state x in time k
%   W_k         [1 x N] Weights for state x in time k

% Your code here!
n = size(X_kmin1,1);
N = size(X_kmin1,2);

for i = 1:N
    X_k(:,i) = proc_f(X_kmin1(:,i)) + mvnrnd(zeros(n,1),proc_Q)';
    ykp =  meas_h(X_k(:,i));

    pyx = exp(-0.5 * ((yk-ykp)' * (meas_R)^-1 * (yk-ykp)));
    W_k(i) = W_kmin1(i)*pyx;
end

Wt = sum(W_k);
W_k = W_k/Wt;
end

内存不足。输入“帮助内存”作为选项

0 个答案: