已知旧均值旧方差和新的一批数据，计算新均值和新方差代码总结

行业边缘的摸鱼怪

已于 2022-10-18 15:22:12 修改

阅读量874

点赞数

分类专栏：代码总结文章标签：均值算法 python

于 2022-10-18 15:15:21 首次发布

本文链接：https://blog.csdn.net/dream6985/article/details/127389265

版权

代码总结专栏收录该内容

11 篇文章 0 订阅

订阅专栏

数学原理

略

代码

def calculate_mean_var(sample_count, new_sample_num, new_history, old_mean, old_var):
    # sample_count：旧样本总数
    # new_sample_num：新样本总数
    # new_history：新的样本
    # old_mean：旧均值
    # old_var：旧方差
    new_mean = np.mean(new_history, axis=0)

    new_var = np.sum(np.square(new_history - new_mean), axis=0)
    new_var = (sample_count * np.square(old_var) + new_var)
    new_var /= (new_sample_num + sample_count)
    new_var = np.sqrt(new_var)

    new_mean = (sample_count * old_mean + new_sample_num * new_mean)
    new_mean /= (new_sample_num + sample_count)

    return new_mean, new_var