统计学习方法第二十章作业：潜在狄利克雷分配 LDA 吉布斯抽样法算法 代码实现

潜在狄利克雷分配 LDA 吉布斯抽样法算法

import numpy as np import jiebaclass LDA:def __init__(self,text_list,k):self.k = kself.text_list = text_listself.text_num = len(text_list)self.get_X()self.NKV = np.zeros((self.k,self.word_num))self.NMK = np.zeros((self.text_num,self.k))self.nm = np.zeros(self.text_num)self.nk = np.zeros(self.k)self.zmn = [[] for i in range(self.text_num)]self.alpha = np.random.randint(1,self.k,size=k)self.beta = np.random.randint(1,self.word_num, size=self.word_num)def get_X(self):self.cuted_text = [jieba.lcut(text,cut_all=True) for text in self.text_list]self.word_all = []for i in self.cuted_text:self.word_all.extend(i)self.word_set = list(set(self.word_all))self.word_num = len(self.word_set)self.word_dict = {}for index,word in enumerate(self.word_set):self.word_dict[word] = indexdef initial_K(self):for doc_num in range(self.text_num):for word in self.cuted_text[doc_num]:k = np.random.choice(self.k, 1)[0]self.zmn[doc_num].append(k)v = self.word_dict[word]self.NMK[doc_num,k] += 1self.nm[doc_num] += 1self.NKV[k,v] += 1self.nk[k] += 1def iter_jbs(self):for doc_num in range(self.text_num):for word_index in range(len(self.cuted_text[doc_num])):v = self.word_dict[self.cuted_text[doc_num][word_index]]k = self.zmn[doc_num][word_index]self.NMK[doc_num,k] -= 1self.nm[doc_num] -= 1self.NKV[k,v] -= 1self.nk[k] -= 1p_klist = (self.NKV[:,v]+self.beta[v])/np.sum(self.NKV[:,v]+self.beta[v])*(self.NMK[doc_num]+self.alpha[k])/np.sum(self.NMK[doc_num]+self.alpha[k])p_klist = p_klist/np.sum(p_klist)k_choice = np.random.choice(self.k,p = p_klist)self.zmn[doc_num][word_index] = k_choiceself.NMK[doc_num,k_choice] += 1self.nm[doc_num] += 1self.NKV[k_choice,v] += 1self.nk[k_choice] += 1def get_sita_y(self):self.sita_mk = np.zeros((self.text_num,self.k))self.yta_kv = np.zeros((self.k,self.word_num))for i in range(self.text_num):self.sita_mk[i] = (self.NMK[i]+self.alpha)/np.sum(self.NMK[i])for j in range(self.k):self.yta_kv[j] = (self.NKV[j]+self.beta)/np.sum(self.NKV[j])def fit(self,max_iter = 100):self.initial_K()for iter in range(max_iter):print(iter)self.iter_jbs()self.get_sita_y()def main():text_list = ['一个月前，足协杯十六进八的比赛，辽足费尽周折对调主客场，目的只是为了葫芦岛体育场的启用仪式。那场球辽足5比0痛宰“主力休息”的天津泰达。几天后中超联赛辽足客场对天津，轮到辽足“全替补”，\1比3输球，甘为天津泰达保级的祭品。那时，辽足以“联赛保级问题不大，足协杯拼一拼”作为主力和外援联赛全部缺阵的理由。','被一脚踹进“忘恩负义”坑里的孙杨，刚刚爬出来，又有手伸出来，要把孙杨再往坑里推。即使是陪伴孙杨参加世锦赛的张亚东(微博)教练，\也没敢大义凛然地伸出援手，“孙杨愿意回去我不拦”，球又踢给了孙杨。张亚东教练怕什么呢？','孙杨成绩的利益分配，以及荣誉的分享，圈里人都知道，拿了世界冠军和全运冠军，运动员都会有相应的高额奖金，那么主管教练也会得到与之对应的丰厚奖励，\所以孙杨获得的荣誉，也会惠及主管教练。']k = 2lda = LDA(text_list,k)lda.fit()print(lda.sita_mk)print(lda.yta_kv)if __name__ == '__main__':main()#result--------------------------[[0.20689655 0.81034483][0.7 0.32222222][0.50666667 0.52 ]][[1.2295082 0.12295082 0.58196721 1.21311475 1.08196721 0.065573771.12295082 0.18032787 0.98360656 0.16393443 0.78688525 1.016393440.7704918 1.12295082 1.01639344 0.43442623 1.00819672 0.721311480.70491803 0.21311475 0.78688525 0.14754098 0.6147541 0.532786890.59836066 1.20491803 0.6557377 0.01639344 1.05737705 0.532786891.22131148 0.71311475 1.29508197 1.23770492 0.59016393 1.204918030.13114754 0.04918033 0.99180328 0.93442623 1.27868852 1.15573770.90983607 0.66393443 1.08196721 1.07377049 0.57377049 0.081967210.17213115 0.54098361 1.14754098 0.98360656 0.17213115 0.262295080.6557377 1.12295082 0.80327869 0.77868852 1.10655738 0.819672130.79508197 0.41803279 0.63934426 0.36065574 1.29508197 0.745901640.99180328 1.14754098 0.67213115 0.33606557 0.40163934 0.737704920.67213115 0.86885246 0.18852459 0.17213115 0.75409836 0.336065570.07377049 1.13114754 0.40163934 0.63934426 0.36885246 1.278688521.19672131 0.35245902 1.10655738 0.21311475 1.19672131 0.713114750.29508197 0.67213115 1.02459016 0.87704918 0.81147541 1.049180330.1147541 1.1147541 0.40163934 1.05737705 0.31147541 0.409836070.31147541 0.59016393 0.74590164 1.18852459 1.32786885 0.745901640.48360656 0.42622951 0.8442623 1.22131148 0.95901639 0.696721310.09836066 1.26229508 1.1147541 0.63934426 1.1557377 0.147540981.18032787 0.1557377 0.93442623 0.63114754 0.45901639 0.524590161.28688525 1.13114754 0.91803279 1.27868852 0.82786885 0.311475410.33606557 0.41803279 1.30327869 0.99180328 1.31147541 1.172131150.97540984 1.19672131 0.24590164 0.90983607 0.59016393 0.491803280.87704918 1.08196721 0.42622951 0.27868852 0.49180328 0.696721310.08196721 0.48360656 0.5 0.7704918 0.95081967 1.0.52459016 0.16393443 1.1147541 0.18852459 0.82786885 1.090163930.1147541 0.93442623][0.99371069 0.10691824 0.44025157 0.93710692 0.83647799 0.056603770.88679245 0.14465409 0.76100629 0.1509434 0.61006289 0.786163520.59119497 0.87421384 0.77987421 0.35220126 0.77358491 0.559748430.53459119 0.17610063 0.60377358 0.11949686 0.47798742 0.408805030.45283019 0.93081761 0.50314465 0.00628931 0.81761006 0.415094340.93081761 0.55345912 0.98742138 0.94339623 0.44654088 0.918238990.10691824 0.03144654 0.7672956 0.72327044 0.98742138 0.880503140.71069182 0.51572327 0.83647799 0.81761006 0.44654088 0.056603770.12578616 0.40880503 0.87421384 0.74842767 0.12578616 0.194968550.50943396 0.85534591 0.62264151 0.60377358 0.8427673 0.660377360.6163522 0.32704403 0.48427673 0.27044025 1.01257862 0.566037740.74213836 0.86792453 0.50943396 0.25157233 0.31446541 0.572327040.52830189 0.65408805 0.1509434 0.12578616 0.58490566 0.264150940.05660377 0.8490566 0.31446541 0.50314465 0.27044025 0.987421380.9245283 0.27672956 0.8427673 0.16981132 0.9245283 0.547169810.2327044 0.52201258 0.77987421 0.67924528 0.63522013 0.798742140.08805031 0.8490566 0.29559748 0.81761006 0.24528302 0.320754720.25157233 0.4591195 0.57861635 0.90566038 1.02515723 0.566037740.36477987 0.32075472 0.65408805 0.93710692 0.72955975 0.528301890.08176101 0.97484277 0.8490566 0.49685535 0.89308176 0.119496860.89937107 0.11320755 0.70440252 0.49056604 0.33962264 0.408805030.98113208 0.86163522 0.69811321 0.97484277 0.66037736 0.245283020.25157233 0.32075472 0.99371069 0.75471698 1.01257862 0.893081760.75471698 0.93081761 0.18238994 0.6918239 0.4591195 0.364779870.66666667 0.80503145 0.32075472 0.22012579 0.3836478 0.528301890.05660377 0.37735849 0.37735849 0.58490566 0.74213836 0.773584910.40251572 0.13207547 0.8490566 0.1509434 0.64150943 0.836477990.09433962 0.72327044]]

总结

以上是生活随笔为你收集整理的统计学习方法第二十章作业：潜在狄利克雷分配 LDA 吉布斯抽样法算法 代码实现的全部内容，希望文章能够帮你解决所遇到的问题。

如果觉得生活随笔网站内容还不错，欢迎将生活随笔推荐给好友。