参考:https://cloud.tencent.com/developer/article/1530432
n_components : 目标输出维度
n_iter :迭代次数
from sklearn.decomposition import TruncatedSVD
lsa = TruncatedSVD(n_components=k, n_iter=n_iters)
M_reduced = lsa.fit_transform(M)
使用TruncatedSVD,把原先规模为(文本数,词汇数)的特征矩阵X化为规模为(文本数,主题数)的新特征矩阵X2:
(主题数一般比词汇数少)