最美情侣中文字幕电影,在线麻豆精品传媒,在线网站高清黄,久久黄色视频

歡迎光臨散文網(wǎng) 會(huì)員登陸 & 注冊

2023阿里巴巴全球數(shù)學(xué)競賽預(yù)選賽題/決賽部分題個(gè)人解 (三)

2023-06-25 19:20 作者:saqatl  | 我要投稿

預(yù)選賽題 7.


這是最優(yōu)停止問題,但是比賽的時(shí)候忙著趕論文,題都沒看就潤了。下面直接用動(dòng)態(tài)規(guī)劃來解決一般的?p%20%5Cin%20(0%2C1)。

設(shè)?m?為正在考慮的候選者,S_m?為其排名的隨機(jī)變量,X_m?為該候選者是否接受 offer 的隨機(jī)變量。所有策略狀態(tài)都可以用?(m%2Cs%2Cx)?表示,其中?s?為?m?的排名,x?表示候選者是否接受 offer。

顯然,若?s%20%5Cne%201r?就不會(huì)被選為最優(yōu)者,此時(shí)應(yīng)該考慮?(m%2B1%2Cs%2Cx)。設(shè)?P(m%2Cs%2Cx)?是狀態(tài)?(m%2Cs%2Cx)?下選到最優(yōu)者的最大概率,則

P(m%2Cs%2Cx)%20%3D%20%5Cmax%5C%7B%5Ctilde%20P(m%2Cs%2Cx)%2C%20%5Cmathbb%7BE%7D(P(m%2B1%2C%20S_%7Bm%2B1%7D%2C%20X_%7Bm%2B1%7D))%5C%7D%20(m%20%3C%20N)

其中?%5Ctilde%20P(m%2Cs%2Cx)?是狀態(tài)?(m%2Cs%2Cx)?即為最優(yōu)者的概率,顯然當(dāng)?s%3D1%2C%20x%3D1?時(shí)?%5Ctilde%7BP%7D(m%2Cs%2Cx)%20%3D%20m%2FN,否則?%5Ctilde%20P(m%2Cs%2Cx)%20%3D%200。當(dāng)?m%20%3D%20N?時(shí),問題的邊界條件為當(dāng)?s%20%3D%201%2C%20x%20%3D%201?時(shí)?P(N%2Cs%2Cx)%20%3D%201,否則?P(N%2Cs%2Cx)%20%3D%200。

現(xiàn)在尋找?%5Cmathbb%7BE%7D(P(m%2CS_m%2CX_m))?應(yīng)當(dāng)滿足的遞推關(guān)系(注意它和?s?無關(guān))。

%5Cbegin%7Baligned%7D%0A%0A%5Cmathbb%7BE%7D(P(m%2CS_m%2CX_m))%20%26%3D%20%5Cfrac%7B1%7D%7Bm%7D%20%5Csum%5Climits_%7Bs%3D1%7D%5Em%20(pP(m%2Cs%2C1)%20%2B%20(1-p)P(m%2Cs%2C0))%5C%5C%0A%0A%26%3D%20%5Cfrac%7Bp%7D%7Bm%7D%20%5Csum%5Climits_%7Bs%3D1%7D%5E%7Bm%7D%20%5Cmax%5C%7B%5Ctilde%20P(m%2Cs%2C1)%2C%20%5Cmathbb%7BE%7D(P(m%2B1%2C%20S_%7Bm%2B1%7D%2C%20X_%7Bm%2B1%7D))%5C%7D%5C%5C%0A%0A%26%5C%20%2B%20%5Cfrac%7B1-p%7D%7Bm%7D%20%5Csum%5Climits_%7Bs%3D1%7D%5E%7Bm%7D%20%5Cmax%5C%7B%5Ctilde%20P(m%2Cs%2C0)%2C%20%5Cmathbb%7BE%7D(P(m%2B1%2C%20S_%7Bm%2B1%7D%2C%20X_%7Bm%2B1%7D))%5C%7D%20%5C%5C%0A%0A%26%3D%20%5Cfrac%7Bp%7D%7Bm%7D((m-1)%5Cmathbb%7BE%7D(P(m%2B1%2C%20S_%7Bm%2B1%7D%2C%20X_%7Bm%2B1%7D))%20%2B%20P(m%2C1%2C1))%20%2B%20%5Cfrac%7B1-p%7D%7Bm%7D%20m%5Cmathbb%7BE%7D(P(m%2B1%2C%20S_%7Bm%2B1%7D%2C%20X_%7Bm%2B1%7D))%20%5C%5C%0A%0A%26%3D%20%5Cfrac%7Bm-p%7D%7Bm%7D%20%5Cmathbb%7BE%7D(P(m%2B1%2C%20S_%7Bm%2B1%7D%2C%20X_%7Bm%2B1%7D))%20%2B%20%5Cfrac%7Bp%7D%7Bm%7D%20P(m%2C1%2C1)%0A%0A%5Cend%7Baligned%7D

當(dāng)?P(m%2C1%2C1)%20%3E%20m%2FN?時(shí),繼續(xù)考慮第?m%2B1?個(gè)候選人所獲得的概率比直接選擇第?m?個(gè)候選人所獲得的概率更大,此時(shí)最優(yōu)策略應(yīng)當(dāng)為考慮第?m%2B1?個(gè)候選人,因此?P(m%2C1%2C1)%20%3D%20%5Cmathbb%7BE%7D(P(m%2B1%2C%20S_%7Bm%2B1%7D%2C%20X_%7Bm%2B1%7D))。此時(shí)代入上式可得?%5Cmathbb%7BE%7D(P(m%2CS_m%2CX_m))%20%3D%20%5Cmathbb%7BE%7D(P(m%2B1%2C%20S_%7Bm%2B1%7D%2C%20X_%7Bm%2B1%7D))%20%3E%20m%2FN%20%3E%20(m-1)%2FN,代入遞推式可得?P(m-1%2C1%2C1)%20%3E%20(m-1)%2FN。因此對第?m-1?個(gè)候選者而言,也應(yīng)當(dāng)繼續(xù)考慮第?m?個(gè)候選人。

因此,最優(yōu)策略滿足如下條件:如果第?m?個(gè)候選人不會(huì)被選,則第?m-1?個(gè)候選者也不會(huì)被選;如果第?m?個(gè)候選人可以被考慮,那么第?m%2B1?個(gè)候選人也應(yīng)當(dāng)被考慮。

如果想在?(m%2C1%2C1)?狀態(tài)選擇第?m?個(gè)候選人,則?P(m%2C1%2C1)%20%3D%20m%2FN,因此

%5Cmathbb%7BE%7D(P(m%2CS_m%2CX_m))%20%3D%20%5Cfrac%7Br-p%7D%7Br%7D%20%5Cmathbb%7BE%7D(P(m%2B1%2C%20S_%7Bm%2B1%7D%2C%20X_%7Bm%2B1%7D))%20%2B%20%5Cfrac%7Bp%7D%7BN%7D%20(m%20%5Cge%20m_N)

其中?m_N?為使得?P(m_N%2C1%2C1)%20%3D%20m_N%2FN?的最小?m?值。現(xiàn)在代入邊界條件,就得到

%5Cmathbb%7BE%7D(P(m%2B1%2C%20S_%7Bm%2B1%7D%2C%20X_%7Bm%2B1%7D))%20%3D%20%5Cfrac%7Bpm%7D%7B(1-p)N%7D%20%5Cleft(%5Cprod%5Climits_%7Bk%3Dm%7D%5E%7BN-1%7D%20%5Cleft(1%20%2B%20%5Cfrac%7B1-p%7D%7Bk%7D%5Cright)%20-%201%5Cright)%20(m%20%5Cge%20m_N)

綜上所述,我們只需要考慮狀態(tài)?(m%2C1%2C1),此時(shí)選擇第?m?個(gè)候選人當(dāng)且僅當(dāng)?%5Cmathbb%7BE%7D(P(m%2B1%2C%20S_%7Bm%2B1%7D%2C%20X_%7Bm%2B1%7D))%20%5Cle%20m%2FN。設(shè)?m_N?是第一個(gè)滿足上述條件的?m,則最優(yōu)策略即為從第?m_N?個(gè)候選人開始考慮發(fā) offer,選擇第一個(gè)?s%20%3D%201?的候選人。

下面求出?m_N。這也即尋找

%5Cprod%5Climits_%7Bk%3Dm_N%7D%5E%7BN-1%7D%20%5Cleft(1%20%2B%20%5Cfrac%7B1-p%7D%7Bk%7D%5Cright)%20%5Cle%20%5Cfrac%7B1%7D%7Bp%7D%2C%20%5Cquad%20%5Cprod%5Climits_%7Bk%3Dm_N%20-%201%7D%5E%7BN-1%7D%20%5Cleft(1%20%2B%20%5Cfrac%7B1-p%7D%7Bk%7D%5Cright)%20%5Cge%20%5Cfrac%7B1%7D%7Bp%7D

由?%5Cmathrm%7BBernoulli%7D?不等式,

%5Cprod%5Climits_%7Bk%3Dm_N%7D%5E%7BN-1%7D%20%5Cleft(1%20%2B%20%5Cfrac%7B1-p%7D%7Bk%7D%5Cright)%20%3E%20%5Cprod%5Climits_%7Bk%3Dm_N%7D%5E%7BN-1%7D%20%5Cleft(1%20%2B%20%5Cfrac%7B1%7D%7Bk%7D%5Cright)%5E%7B1-p%7D%20%3D%20%5Cleft(%5Cfrac%7BN%7D%7Bm_N%7D%5Cright)%5E%7B1-p%7D

另一方面,

%5Cbegin%7Baligned%7D%0A%0A%5Cfrac%7B1%7D%7B%5Cprod%5Climits_%7Bk%3Dm_N%7D%5E%7BN-1%7D%20%5Cleft(1%20%2B%20%5Cfrac%7B1-p%7D%7Bk%7D%5Cright)%7D%20%26%3D%20%5Cprod%5Climits_%7Bk%3Dm_N%7D%5E%7BN-1%7D%20%5Cleft(%5Cfrac%7Bk%7D%7Bk%2B1-p%7D%5Cright)%20%3D%20%5Cprod%5Climits_%7Bk%3Dm_N%7D%5E%7BN-1%7D%20%5Cleft(1%20-%20%5Cfrac%7B1-p%7D%7Bk%2B1-p%7D%5Cright)%20%5C%5C%0A%0A%26%3E%20%5Cprod%5Climits_%7Bk%3Dm_N%7D%5E%7BN-1%7D%20%5Cleft(1%20%2B%20%5Cfrac%7B1%7D%7Bk%2B1-p%7D%5Cright)%5E%7B1-p%7D%20%3D%20%5Cprod%5Climits_%7Bk%3Dm_N%7D%5E%7BN-1%7D%20%5Cleft(%5Cfrac%7Bk-p%7D%7Bk%2B1-p%7D%5Cright)%0A%0A%5Cend%7Baligned%7D

%5Cprod%5Climits_%7Bk%3Dm_N%7D%5E%7BN-1%7D%20%5Cleft(1%20%2B%20%5Cfrac%7B1-p%7D%7Bk%7D%5Cright)%20%3C%20%5Cprod%5Climits_%7Bk%3Dm_N%7D%5E%7BN-1%7D%20%5Cleft(%5Cfrac%7Bk%2B1-p%7D%7Bk-p%7D%5Cright)%20%3D%20%5Cleft(%5Cfrac%7BN-p%7D%7Bm_N%20-%20p%7D%5Cright)%5E%7B1-p%7D

因此

%5Cleft(%5Cfrac%7BN%7D%7Bm_N%7D%5Cright)%5E%7B1-p%7D%20%3C%20%5Cfrac%7B1%7D%7Bp%7D%20%3C%20%5Cleft(%5Cfrac%7BN-p%7D%7Bm_N%20-%201%20-%20%20p%7D%5Cright)%5E%7B1-p%7D

p%5E%7B%5Cfrac%7B1%7D%7B1-p%7D%7D%20%3C%20%5Cfrac%7Bm_N%7D%7BN%7D%20%3C%20%5Cfrac%7BN-p%7D%7BN%7D%20p%5E%7B%5Cfrac%7B1%7D%7B1-p%7D%7D%20%2B%20%5Cfrac%7Bp%2B1%7D%7BN%7D

兩側(cè)同取極限得

%5Clim%5Climits_%7BN%20%5Cto%20%5Cinfty%7D%20%5Cfrac%7Bm_N%7D%7BN%7D%20%3D%20p%5E%7B%5Cfrac%7B1%7D%7B1-p%7D%7D

當(dāng)?p%20%3D%201?時(shí),上述極限即趨于?1%2Fe

決賽我選的主分析副應(yīng)用與計(jì)算?,F(xiàn)在看來選得還行。

分析題 1.?數(shù)列?%5C%7Ba_n%5C%7D?滿足

a_%7Bn%2B1%7D%20%3D%20a_n%20%2B%20%5Cfrac%7Ba_n%5E2%7D%7Bn%5E2%7D%2C%20%5Cquad%20a_1%20%3D%20%5Cfrac%7B2%7D%7B5%7D

證明:對任意正整數(shù)?n?都有?a_n%20%3C%201。

沒什么想法,就硬估計(jì)的。自然會(huì)想到利用歸納法,

a_%7Bn%2B1%7D%20%3D%20%5Csum%5Climits_%7Bi%3D1%7D%5En%20(a_%7Bi%2B1%7D%20-%20a_i)%20%2B%20a_1%20%3D%20%5Csum%5Climits_%7Bi%3D1%7D%5En%20%5Cfrac%7Ba_i%5E2%7D%7Bi%5E2%7D%20%2B%20a_1

再用?%5Cmathrm%7BBessel%7D?問題放縮。不過從?a_1?開始不太行,計(jì)算到?a_4%20%5Capprox%200.68?之后就可以了:

a_%7Bn%2B1%7D%20%3D%20%5Csum%5Climits_%7Bi%3D4%7D%5En%20%5Cfrac%7Ba_i%5E2%7D%7Bi%5E2%7D%20%2B%20a_4%20%3C%20%5Csum%5Climits_%7Bi%3D4%7D%5En%20%5Cfrac%7Ba_i%5E2%7D%7Bi%5E2%7D%20%2B%20a_4%20%3C%20%5Cfrac%7B%5Cpi%5E2%7D%7B6%7D%20-%201%20-%20%5Cfrac%7B1%7D%7B4%7D%20-%20%5Cfrac%7B1%7D%7B9%7D%20%2B%20a_4%20%5Capprox%200.97%20%3C%201

分析題 2.?設(shè)?V?是平面閉區(qū)域?Q%20%3D%20%5B0%2C1%5D%5E2?上由連續(xù)函數(shù)構(gòu)成的?n?維線性空間,證明:存在函數(shù)?f%20%5Cin%20V?使得

%5C%7Cf%5C%7C_%7BL%5E2(Q)%7D%20%3D%201%2C%20%5Cquad%20%5C%7Cf%5C%7C_%7BL%5E%7B%5Cinfty%7D(Q)%7D%20%5Cge%20%5Csqrt%7Bn%7D

取?L%5E2?范數(shù)下的標(biāo)準(zhǔn)正交基?f_1%2C%20f_2%2C%20%5Cldots%2C%20f_n。則對任意?f%20%5Cin%20V,有?f%20%3D%20%5Csum%5Climits_%7Bi%3D1%7D%5En%20%5Calpha_i%20f_i。

先考慮讓?%5C%7Cf%5C%7C_%7BL%5E2(Q)%7D%20%3D%201,不難知道?%5C%7Cf%5C%7C_%7BL%5E2(Q)%7D%20%3D%20%5Csqrt%7B%5Csum%5Climits_%7Bi%3D1%7D%5En%20%5Calpha_i%5E2%7D,故?%5Csum%5Climits_%7Bi%3D1%7D%5En%20%5Calpha_i%5E2%20%3D%201。

另一方面,由于?%5Csum%5Climits_%7Bi%3D1%7D%5En%20%5C%7Cf_i%5C%7C%5E2_%7BL%5E2(Q)%7D%20%3D%20n,則由積分中值定理可知存在?%5Cxi%20%5Cin%20Q?使得?%5Csum%5Climits_%7Bi%3D1%7D%5En%20f_i%5E2(%5Cxi)%20%3D%20n。此時(shí)由?%5Cmathrm%7BCauchy%7D?不等式可知?f(%5Cxi)%20%3D%20%5Csum%5Climits_%7Bi%3D1%7D%5En%20%5Calpha_i%20f_i(%5Cxi)%20%5Cle%20%5Csqrt%7B%5Csum%5Climits_%7Bi%3D1%7D%5En%20%5Calpha_i%5E2%7D%20%5Csqrt%7B%5Csum%5Climits_%7Bi%3D1%7D%5En%20f_i%5E2(%5Cxi)%7D%20%3D%20%5Csqrt%7Bn%7D,且等號(hào)顯然可以取到。此時(shí)的?f?即滿足題目條件。

總感覺以前做過這題,但是翻了翻還存著的 TD 沒找著。

2023阿里巴巴全球數(shù)學(xué)競賽預(yù)選賽題/決賽部分題個(gè)人解 (三)的評(píng)論 (共 條)

分享到微博請遵守國家法律
岳阳市| 肇庆市| 台江县| 通江县| 瑞昌市| 霍邱县| 永清县| 泸西县| 孟津县| 昆山市| 洛川县| 石门县| 曲麻莱县| 邵阳县| 加查县| 潍坊市| 庄浪县| 通许县| 贵港市| 呼图壁县| 汤阴县| 新宾| 遂昌县| 习水县| 南开区| 乌拉特中旗| 彭泽县| 博野县| 沁水县| 泊头市| 抚松县| 青冈县| 辉县市| 临沭县| 忻城县| 中江县| 宜兰县| 信丰县| 喀喇沁旗| 方城县| 阜康市|