编写具体的代码

好的，下面是一个可能的实现方法：

% 读取csv文件
spectra = readtable('spectra.csv');

% 提取数据矩阵
data = table2array(spectra(2:end, 2:end));

% 定义适应度函数
% 这里使用简单的knn分类器，分类精度作为适应度
function [accuracy] = fitness_func(selected_wavelengths, data)
    % 转换二进制编码为波段数值
    wavelengths = find(selected_wavelengths);
    % 选择特征波段
    selected_data = data(:, wavelengths);
    % 分割数据集
    cv = cvpartition(size(selected_data, 1), 'HoldOut', 0.3);
    train_data = selected_data(training(cv), :);
    test_data = selected_data(test(cv), :);
    train_labels = spectra(training(cv), 1);
    test_labels = spectra(test(cv), 1);
    % 训练knn分类器
    mdl = fitcknn(train_data, train_labels);
    % 预测测试集
    pred_labels = predict(mdl, test_data);
    % 计算分类精度
    accuracy = sum(test_labels == pred_labels) / numel(test_labels);
end

% 定义遗传算法参数
pop_size = 50;
num_vars = size(data, 2);
fitness_func_handle = @(selected_wavelengths) fitness_func(selected_wavelengths, data);
options = gaoptimset('PopulationSize', pop_size, ...
                     'Generations', 50, ...
                     'CrossoverFraction', 0.8, ...
                     'MutationFcn', {@mutationuniform, 0.1}, ...
                     'Display', 'iter', ...
                     'PlotFcn', @gaplotbestf);

% 运行遗传算法
[selected_wavelengths, max_fitness] = ga(fitness_func_handle, num_vars, options);

% 输出选择的特征波段
selected_wavelengths_idx = find(selected_wavelengths);
selected_wavelengths_values = spectra.Properties.VariableNames(selected_wavelengths_idx + 1);
fprintf('Selected wavelengths: %s\n', strjoin(selected_wavelengths_values, ','));

% 保存选择的特征波段数据
selected_data = data(:, selected_wavelengths_idx);
selected_spectra = spectra(:, [1, selected_wavelengths_idx + 1]);
writetable(selected_spectra, 'selected_spectra.csv');

需要注意的是，由于没有提供数据集，上述代码未经测试，可能存在一些错误。如果出现问题，可以根据错误提示进行调试。