Bayesian Models

Bayesian Models#

Probabilistic and ranking-oriented matrix factorization methods.

Sandbox only

All Bayesian models in CoreRec are sandbox / experimental. There is no production-tier Bayesian model in the 14-model CI suite.

Sandbox models#

Model

Import

Tutorial

BPR

from corerec.sandbox.collaborative_full.cornac_bpr import BPR

BPR

BPR-MF

from corerec.sandbox.collaborative_full.bayesian_method_base.bprmf_base import BPRMF_base

BPR-MF

VMF

from corerec.sandbox.collaborative_full.mf_base.vmf_base import VMF_base

VMF

Example (BPR)#

from corerec.sandbox.collaborative_full.cornac_bpr import BPR

model = BPR(k=50, max_iter=100, learning_rate=0.01, lambda_reg=0.01, verbose=True)
model.fit(user_ids, item_ids)  # implicit feedback pairs

Legacy paths

Do not use corerec.engines.unionizedFilterEngine.* or corerec.engines.bpr — those modules do not exist. Import from corerec.sandbox.* as shown in each tutorial.

When to use#

  • Implicit feedback (clicks, views)

  • Pairwise ranking optimization (BPR)

  • Research prototypes requiring uncertainty (VMF)

See also#

Contents