当前位置:网站首页>极大似然估计
极大似然估计
2022-08-02 02:20:00 【wekings】
一.极大似然估计
通俗定义:已知样本的结果信息,求在哪种模型下,出现这些样本结果的概率最大。极大似然估计法提供了一种用观测数据来评估模型参数的方法。
举例说明:
ex.袋子中有黑白两种球,颜色比例不知,做100次有放回抽样(独立同分布),其中70次为白球,30次为黑球,极大似然估计就可通过样本的抽样结果求得袋子中球的分布模型的参数。
P(样本结果|Model)=P(x1,x2,x3,...,x100|Model)
多个相互独立事件同时发生的概率=每个事件发生概率的乘积,因此,
上式=P(x1|Model)P(x2|Model)...P(x100|Model)=
p就是要求的模型参数值,不同的p导致P(样本结果|Model)不同,有无数种选择,极大似然估计要选择使P(样本结果|Model)值最大的参数p,将其看作p的函数,则令其导数=0,即可求得使P(样本结果|Model)最大得参数值p。
,求得p=0.7。
二.李航统计学习中:当模型是条件概率分布,损失函数是对数函数时,经验风险最小化就等价于极大似然估计
假设模型条件概率分布为:
最小化经验风险 :
等价于:
极大似然估计:
将式2取对数后累乘变累加,再求最大值,即为公式1,因此两者等价。
边栏推荐
- LeetCode brushing diary: 33. Search and rotate sorted array
- Can Youxuan database import wrongly be restored?
- 密码学的基础:X.690和对应的BER CER DER编码
- Redis 底层的数据结构
- Reflex WMS Intermediate Series 7: What should I do if I want to cancel the picking of an HD that has finished picking but has not yet been loaded?
- messy website
- Entry name 'org/apache/commons/codec/language/bm/gen_approx_greeklatin.txt' collided
- [LeetCode Daily Question]——654. The largest binary tree
- TKU remembers a single-point QPS optimization (I wish ITEYE is finally back)
- NIO‘s Sword(牛客多校赛)
猜你喜欢
ofstream,ifstream,fstream read and write files
Ringtone 1161. Maximum In-Layer Elements and
BioVendor人俱乐部细胞蛋白(CC16)Elisa试剂盒研究领域
How engineers treat open source
Outsourcing worked for three years, it was abolished...
Power button 1374. Generate each character string is an odd number
Safety (2)
Software testing Interface automation testing Pytest framework encapsulates requests library Encapsulates unified request and multiple base path processing Interface association encapsulation Test cas
FOFAHUB使用测试
LeetCode Brushing Diary: 74. Searching 2D Matrix
随机推荐
Check if IP or port is blocked
¶ Backtop back to the top is not effective
How to adjust the cross cursor too small, CAD dream drawing calculation skills
swift project, sqlcipher3 -> 4, cannot open legacy database is there a way to fix it
Remember a pit for gorm initialization
"NetEase Internship" Weekly Diary (3)
2022-08-01 mysql/stoonedb慢SQL-Q18分析
AI目标分割能力,无需绿幕即可实现快速视频抠图
The Paddle Open Source Community Quarterly Report is here, everything you want to know is here
AOF重写
messy website
Multi-Party Threshold Private Set Intersection with Sublinear Communication-2021: Interpretation
AWR分析报告问题求助:SQL如何可以从哪几个方面优化?
Entry name 'org/apache/commons/codec/language/bm/gen_approx_greeklatin.txt' collided
GTK RGB图像绘制
Ringtone 1161. Maximum In-Layer Elements and
Electronic Manufacturing Warehouse Barcode Management System Solution
The first time I wrote a programming interview question for Niu Ke: input a string and return the letter with the most occurrences of the string
FOFAHUB使用测试
【web】理解 Cookie 和 Session 机制