mode = "offline"
knn = 100
topk = 750
complete_name = "maurizio_" + mode + "__knn=" + str(knn) + "_topk=" + str(topk)
if __name__ == '__main__':
sim = sps.load_npz(ROOT_DIR +
"/similarities/offline-similarity_rp3beta_knn100.npz")
dr = Datareader(mode=mode, only_load=True)
######### MAURIZ
ICM = dr.get_icm(alid=True)
cfw = CFW_D_Similarity_Linalg(URM_train=dr.get_urm(),
ICM=ICM.copy(),
S_matrix_target=sim,
URM_validation=None)
cfw.fit()
weights = sps.diags(cfw.D_best)
sps.save_npz("ICM_fw_maurizio", weights)
ICM_weighted = ICM.dot(weights)
sps.save_npz("ICM_fw_maurizio", ICM_weighted)
album = True
artist = False
elif feature_type == 'artist':
album = False
artist = True
else:
print("invalid type")
complete_name = mode + "_" + name + "_" + feature_type + "_knn=" + str(
knn) + "_topk=" + str(topk)
if mode == "offline":
dr = Datareader(verbose=False, mode=mode, only_load=True)
urm = dr.get_urm()
icm = dr.get_icm(arid=artist, alid=album)
pid = dr.get_test_pids()
icm_bm25 = pre.bm25_row(icm)
cbfi = Knn_content_item()
cbfi.fit(urm, icm_bm25, pid)
cbfi.compute_model(top_k=knn,
sm_type=TVERSKY,
shrink=100,
alpha=0.1,
binary=False,
verbose=True)
cbfi.compute_rating(top_k=topk, verbose=True, small=True)
sps.save_npz(complete_name + ".npz", cbfi.eurm)
import numpy as np
import sys
from utils.datareader import Datareader
from utils.definitions import *
from utils.evaluator import Evaluator
from sklearn.utils.sparsefuncs import inplace_csr_column_scale
from recommenders.similarity.dot_product import dot_product
from recommenders.similarity.s_plus import tversky_similarity
import time
from utils.post_processing import *
from scipy import sparse
from utils.pre_processing import *
from utils.submitter import Submitter
from fbpca import pca
from fbpca import set_matrix_mult
datareader = Datareader(mode='offline', only_load=True)
evaluator = Evaluator(datareader)
urm = datareader.get_urm()
icm = datareader.get_icm(arid=True)
print('PCA...')
u, s, v = pca(icm, k=100)
print('Dot...')
icm_new = set_matrix_mult()
print(icm_new)
trs = df['tid'].values
n = len(df)
pids = df['new_pid'].values
del df
n_pids = pids.max() + 1 # index starts from 0
# create partial icm
icm_ = sp.csr_matrix((np.ones(n), (trs, pids)),
shape=(n_tracks, n_pids),
dtype=np.int32)
icm_cat8 = sp.hstack([icm_cat8, icm_])
icm_cat10 = sp.hstack([icm_cat10, icm_])
urm = dr.get_urm()
icm_pl = urm.copy().T
icm_al = dr.get_icm(arid=False, alid=True)
icm_ar = dr.get_icm(arid=True, alid=False)
icm_cat8 = sp.hstack([icm_cat8, icm_al, icm_ar])
icm_cat10 = sp.hstack([icm_cat10, icm_al, icm_ar])
# hybrid cat8
for i in range(0, 5):
icm_cat8 = sp.hstack([icm_cat8, icm_pl])
# hybrid cat10
for i in range(0, 3):
icm_cat10 = sp.hstack([icm_cat10, icm_pl])
# dump the icms
sp.save_npz(ROOT_DIR + '/data/hybrid_icm_cat8_' + mode + '.npz', icm_cat8)
