《Selection of Test Samples to Improve DNN Test Efficiency based on Neuron Clusters》

思路:

1.预测的设定:与训练集的数据A 激活神经元相似的种子X,就被预测为A类。

        若种子X被DNN分为A类,则激活神经元应相似于A。若DNN将X错误分类为B,激活神经元也要相似于B。

实验结果:用皮尔逊相关系数分析

2.当比较基于测试样本的神经元簇与基于训练数据集的神经元簇之间的相似度时,如果相似度之间的差异小于阈值,则将测试样本视为故障诱导测试样本。

  • 0
    点赞
  • 0
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
Sure, here's an example of a multi-carrier system based on Constant Modulus Feedback (CMFB) in MATLAB: ``` %% Define parameters N = 1024; % Number of subcarriers M = 16; % Modulation order P = 4; % Number of pilot subcarriers L = 8; % Length of equalizer sigma2 = 0.1; % Noise variance %% Generate symbols x = randi([0 M-1], N-P, 1); % Random data symbols p = randi([0 M-1], P, 1); % Pilot symbols %% Modulate symbols x_mod = qammod(x, M); % QAM modulation p_mod = qammod(p, M); % QAM modulation %% Add pilot symbols x_with_pilot = zeros(N, 1); x_with_pilot(1:P) = p_mod; x_with_pilot(P+1:end) = x_mod; %% Generate channel coefficients h = randn(L, 1) + randn(L, 1)*1i; % Complex channel coefficients %% Generate received signal y = conv(h, x_with_pilot); % Channel convolution y = y(P+1:end-L+1); % Remove pilot and excess convolution samples y = awgn(y, sigma2); % Add noise %% CMFB equalization H = toeplitz(h, zeros(1, N-L)); e = zeros(N, 1); mu = 0.1; % Step size w = zeros(L, 1); % Initial weights for i = 1:N z = H(:,i)' * w; % Estimate of symbol e(i) = y(i) - z; % Error w = w + mu * e(i) * H(:,i); % Update weights end %% Calculate BER x_hat = qamdemod(H' * w, M); % Demodulate received symbols ber = sum(x ~= x_hat) / length(x); disp(['BER: ' num2str(ber)]); ``` This code generates random QAM symbols, adds pilot symbols, convolves with a random channel impulse response, adds noise, and then uses CMFB equalization to estimate the transmitted symbols. Finally, it calculates the bit error rate (BER) to evaluate the performance of the system.
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值