在Julia中对大型数组中的列集进行二次采样的最快方法

using Base.Threads

function subsample_columns(data::AbstractMatrix, sample_ratio::Float64)
    num_cols = size(data, 2)
    num_sampled_cols = Int(round(num_cols * sample_ratio))
    sampled_cols = Vector{Int}(undef, num_sampled_cols)

    @threads for i in 1:num_sampled_cols
        sampled_cols[i] = rand(1:num_cols)
    end

    sampled_data = similar(data, size(data, 1), num_sampled_cols)

    @threads for i in 1:num_sampled_cols
        sampled_data[:, i] = data[:, sampled_cols[i]]
    end

    return sampled_data, sampled_cols
end

# 示例用法
data = rand(1000, 10000)  # 假设有一个1000行10000列的大型数组
sampled_data, sampled_cols = subsample_columns(data, 0.1)  # 对列集进行10%的二次采样