推荐经典算法实现之BPMF(python+MovieLen)

因前一篇https://blog.youkuaiyun.com/fjssharpsword/article/details/97000479采样问题未解决，发现如下github上有BPMF代码，采用wishart先验，性能和pymc3一致。

参考：https://github.com/LoryPack/BPMF

# coding:utf-8  
'''
@author: Jason.F
@data: 2019.08.01
@function: baseline BPMF(Bayesian Probabilistic Matrix Factorization)
           Datatset: MovieLens-1m:https://grouplens.org/datasets/movielens/  
           Evaluation: RMSE
'''

import numpy as np
import random
import pandas as pd
from numpy.random import multivariate_normal
from scipy.stats import wishart

class DataSet:
    def __init__(self):
        self.trainset, self.testset, self.maxu, self.maxi, self.maxr = self._getDataset_as_list()
        
    def _getDataset_as_list(self):
        #trainset
        filePath = "/data/fjsdata/BMF/ml-1m.train.rating" 
        data = pd.read_csv(filePath, sep='\t', header=None, names=['user', 'item', 'rating'], \
                                 usecols=[0, 1, 2], dtype={0: np.int32, 1: np.int32, 2: np.float})
        maxu, maxi, maxr = data['user'].max()+1, data['item'].max()+1, data['rating'].max()
        print('Dataset Statistics: Interaction = %d, User = %d, Item = %d, Sparsity = %.4f' % \
                  (data.shape[0], maxu, maxi, data.shape[0]/(maxu*maxi)))
        trainset = data.values.tolist()
        #testset
        filePath = "/data/fjsdata/BMF/ml-1m.test.rating" 
        data = pd.read_csv(filePath, sep='\t', header=None, names=['user', 'item', 'rating'], \
                                 usecols=[0, 1, 2], dtype={0: np.int32, 1: np.int32, 2: np.float})
        testset = data.values.tolist()
        return trainset, testset, maxu, maxi, maxr 
    
    def list_to_matrix(self, dataset, maxu, maxi):              
        dataMat = np.zeros([maxu, maxi], dtype=np.float32)
        for u,i,r in dataset:
            dataMat[int(u)][int(i)] = float(r)
        return np.array(dataMat)
    
def Normal_Wishart(mu_0, lamb, W, nu, seed=None):
    """Function extracting a Normal_Wishart random variable"""
    # first draw a Wishart distribution:
    Lambda = wishart(df=nu, scale=W, seed=seed).rvs()  # NB: Lambda is a matrix.
    # then draw a Gaussian multivariate RV with mean mu_0 and(lambda*Lambda)^{-1} as covariance matrix.
    cov = np.linalg.inv(lamb * Lambda)  # this i