Optimum polynomial

If we are presented with the first k terms of a sequence it is impossible to say with certainty the value of the next term, as there are infinitely many polynomial functions that can model the sequence.

As an example, let us consider the sequence of cube numbers. This is defined by the generating function,
un = n3: 1, 8, 27, 64, 125, 216, …

Suppose we were only given the first two terms of this sequence. Working on the principle that “simple is best” we should assume a linear relationship and predict the next term to be 15 (common difference 7). Even if we were presented with the first three terms, by the same principle of simplicity, a quadratic relationship should be assumed.

We shall define OP(k, n) to be the nth term of the optimum polynomial generating function for the first k terms of a sequence. It should be clear that OP(k, n) will accurately generate the terms of the sequence for n ≤ k, and potentially the first incorrect term (FIT) will be OP(k, k+1); in which case we shall call it a bad OP(BOP).

As a basis, if we were only given the first term of sequence, it would be most sensible to assume constancy; that is, for n ≥ 2, OP(1, n) = u1.

Hence we obtain the following OPs for the cubic sequence:

   
OP(1, n) = 1 1, 1, 1, 1, …
OP(2, n) = 7n−6 1, 8, 15, …
OP(3, n) = 6n2−11n+6 1, 8, 27, 58, …
OP(4, n) = n3 1, 8, 27, 64, 125, …

Clearly no BOPs exist for k ≥ 4.

By considering the sum of FITs generated by the BOPs (indicated in red above), we obtain 1 + 15 + 58 = 74.

Consider the following tenth degree polynomial generating function:

un = 1 − n + n2 − n3 + n4 − n5 + n6 − n7 + n8 − n9 + n10

Find the sum of FITs for the BOPs.


最优多项式

如果我们知道了一个数列的前k项,我们仍无法确定地给出下一项的值,因为有无穷个多项式生成函数都有可能是这个数列的模型。

例如,让我们考虑立方数的序列,它可以用如下函数生成,
un = n3: 1, 8, 27, 64, 125, 216, …

如果我们只知道数列的前两项,秉承“简单至上”的原则,我们应当假定这个数列遵循线性关系,并且预测下一项为15(公差为7)。即使我们知道了数列的前三项,根据同样的原则,我们也应当首先假定数列遵循二次函数关系。

给定数列的前k项,定义OP(k, n)是由最优多项式生成函数给出的第n项的值。显然OP(k, n)可以精确地给出n ≤ k的那些项,而可能的第一个不正确项(First Incorrect Term,简记为FIT)将会是OP(k, k+1);如果事实的确如此,我们称这个多项式为坏最优多项式(Bad OP,简记为BOP)。

在最基本的情况下,如果我们只得到了数列的第一项,我们应当假定数列为常数,也就是说,对于n ≥ 2,OP(1, n) = u1。

由此,我们得到了立方数列的最优多项式如下:

   
OP(1, n) = 1 1, 1, 1, 1, …
OP(2, n) = 7n−6 1, 8, 15, …
OP(3, n) = 6n2−11n+6 1, 8, 27, 58, …
OP(4, n) = n3 1, 8, 27, 64, 125, …

显然,当k ≥ 4时不存在坏最优多项式。

所有坏最优多项式的第一个不正确项(用红色标示的数)之和为1 + 15 + 58 = 74。

考虑下面这个十阶多项式生成函数:

un = 1 − n + n2 − n3 + n4 − n5 + n6 − n7 + n8 − n9 + n10

求其所有坏最优多项式的第一个不正确项之和。

解题

mathblog 提到拉格朗日多项式,突然明白了。

wiki中拉格朗日多项式的定义,在数值计算方法中叫拉格朗日差值函数

上面第k+1项就是所求的答案,至于为什么?已知的点很显然能够准确的预测出来,对于未知的点,为什么第k+1个点不能够预测对?

根据上面博客中写的程序,我记忆中本科时候好像写过这个程序的。

Java

package Level4;

public class PE0101{
public static void run(){
Lagrange(); }
public static void Lagrange(){
long[] coef = {1,-1,1,-1,1,-1,1,-1,1,-1,1};
Polynomial poly = new Polynomial(coef); long[] y = new long[coef.length];
for(int i=0;i<y.length;i++)
y[i] = poly.evaluate(i+1); long fits = 0;
for(int n=1;n<=coef.length -1;n++){
long result = 0;
for(int i =1;i<=n;i++){
long tmp1 = 1;
long tmp2 = 1;
for(int j=1;j<=n;j++){
if(i==j)
continue;
else{
tmp1 *= n + 1-j;
tmp2 *= i-j;
}
}
result +=tmp1*y[i-1]/tmp2;
}
fits +=result;
}
System.out.println(fits);
} public static void main(String[] args){
long t0 = System.currentTimeMillis();
run();
long t1 = System.currentTimeMillis();
long t = t1 - t0;
System.out.println("running time="+t/1000+"s"+t%1000+"ms");
// 37076114526
// running time=0s1ms
}
}
class Polynomial{
private long[] coef;
public int Degree;
public Polynomial(int deg){
Degree = deg;
coef = new long[deg+1];
}
public Polynomial(long[] coef){
Degree = coef.length - 1;
this.coef = coef;
}
public long get(int i){
return coef[i];
}
public void set(int i,long value){
coef[i] = value;
}
public long evaluate(long x){
long result =0;
for(int i= this.Degree;i>=0;i--){
result = result *x +get(i);
}
return result;
} }

Python

# coding=gbk

import time as time
import re
import math def run():
y = ploy()
fits = 0
for n in range(1,11):
res = 0
for i in range(1,n+1):
tmp1 = 1
tmp2 = 1
for j in range(1,n+1):
if i==j:
continue
else:
tmp1 *= (n+1-j)
tmp2 *= (i-j)
res += tmp1*y[i-1]/tmp2
fits += res
print res
def ploy():
coef = [1,-1,1,-1,1,-1,1,-1,1,-1,1]
y = list() for n in range(1,11):
res = 1
m = n
for i in range(1,11):
res = res + coef[i] * m
m *=n
y.append(res)
print y
return y t0 = time.time()
run()
t1 = time.time()
print "running time=",(t1-t0),"s"

Project Euler 101 :Optimum polynomial 最优多项式的更多相关文章

  1. Python练习题 048:Project Euler 021:10000以内所有亲和数之和

    本题来自 Project Euler 第21题:https://projecteuler.net/problem=21 ''' Project Euler: Problem 21: Amicable ...

  2. Python练习题 034:Project Euler 006:和平方与平方和之差

    本题来自 Project Euler 第6题:https://projecteuler.net/problem=6 # Project Euler: Problem 6: Sum square dif ...

  3. [project euler] program 4

    上一次接触 project euler 还是2011年的事情,做了前三道题,后来被第四题卡住了,前面几题的代码也没有保留下来. 今天试着暴力破解了一下,代码如下: (我大概是第 172,719 个解出 ...

  4. Python练习题 029:Project Euler 001:3和5的倍数

    开始做 Project Euler 的练习题.网站上总共有565题,真是个大题库啊! # Project Euler, Problem 1: Multiples of 3 and 5 # If we ...

  5. Project Euler 9

    题意:三个正整数a + b + c = 1000,a*a + b*b = c*c.求a*b*c. 解法:可以暴力枚举,但是也有数学方法. 首先,a,b,c中肯定有至少一个为偶数,否则和不可能为以上两个 ...

  6. Project Euler 44: Find the smallest pair of pentagonal numbers whose sum and difference is pentagonal.

    In Problem 42 we dealt with triangular problems, in Problem 44 of Project Euler we deal with pentago ...

  7. project euler 169

    project euler 169 题目链接:https://projecteuler.net/problem=169 参考题解:http://tieba.baidu.com/p/2738022069 ...

  8. 【Project Euler 8】Largest product in a series

    题目要求是: The four adjacent digits in the 1000-digit number that have the greatest product are 9 × 9 × ...

  9. Project Euler 第一题效率分析

    Project Euler: 欧拉计划是一系列挑战数学或者计算机编程问题,解决这些问题需要的不仅仅是数学功底. 启动这一项目的目的在于,为乐于探索的人提供一个钻研其他领域并且学习新知识的平台,将这一平 ...

随机推荐

  1. maven学习手记 - 2

    学习目标 maven 的插件和目标: maven 的生命周期和阶段.   前言 在手记1中看到执行 mvn clean package 时,maven 自动执行了compile 和 test 操作. ...

  2. 1.总结---tr()和QTextCodec对象

    1. 关于Qt 中的tr()函数-------http://tscsh.blog.163.com/blog/static/200320103201310213312518/ 在论坛中漂,经常遇到有人遇 ...

  3. modelsim仿真错误解决办法

    编译不成功可能是因为: 1.本身程序有问题. 2.没有设置顶层文件 3.modelsim 出现错误是不要只是看错误的地方,也要看前面的一部分

  4. js从富文本中找出固定电话、手机号码、邮箱并添加上链接。

    function richtextfindtel(value) { /*固定电话.手机号码.邮箱*/ var tempValue = value.replace(/<[^>]+>/g ...

  5. P1574: [Usaco2009 Jan]地震损坏Damage

    卧槽卧槽卧槽,这道水题竟然让我WA了两遍!!评测系统卡了然后手贱又提交了一次,然后就悲催了呜呜.. 把与不能回家但牛棚完好的牛相邻的牛棚赋值为不能走(false),可以证明,如果该牛回不了家,则周围一 ...

  6. ED/EP简介

    ED:electronic Deposit,电子存折 EP:electronic Purse,电子钱包 PIN:personal identification number,个人识别码 MAC:Mes ...

  7. ios-仿新浪微博app-第1天UI搭建

    1:不用storyboard 点击工程删除main  ui加载全部手码  >> 在application的代理方法didFinishLaunchingWithOptions中添加代码显示w ...

  8. hibernate tool连接oracle生成pojo和xml文件无法查询表解决办法

    需要在hibernate的配置文件中增加 <property name="hibernate.default_schema">[username]</proper ...

  9. 对MVC的理解

    摘要:本文主要谈到了对PHP开发中MVC开发模式的理解. 当用户通过url触发命令时,例如url=http://control.blog.sina.com.cn/admin/article/artic ...

  10. VMM学习-vmm_log

    功能类似verilog里的$display函数,在vmm里做了强化,可以在仿真过程中看到整个平台的运行信息,用来调试仿真平台. 函数原型在vmm.sv里(class vmm_log;),其构造函数为e ...