(
MultVAEOptimizer,
1,
dict(dim_z=200, enc_hidden_dims=600, kl_anneal_goal=0.2
), # nothing to tune, use the parameters used in the paper.
),
# (SLIMOptimizer, 40, dict()), # Note: this is a heavy one.
]
for optimizer_class, n_trials, config in test_configs:
recommender_name = optimizer_class.recommender_class.__name__
optimizer: BaseOptimizer = optimizer_class(
data_train.X_all,
valid_evaluator,
fixed_params=config,
)
(best_param,
validation_result_df) = optimizer.optimize(n_trials=n_trials)
validation_result_df["recommender_name"] = recommender_name
validation_results.append(validation_result_df)
pd.concat(validation_results).to_csv(f"validation_scores.csv")
test_recommender = optimizer.recommender_class(X_train_val_all,
**best_param)
test_recommender.learn()
test_scores = test_evaluator.get_scores(test_recommender,
[20, 50, 100])
test_results.append(
dict(name=recommender_name, best_param=best_param, **test_scores))
with open("test_results.json", "w") as ofs:
json.dump(test_results, ofs, indent=2)
(TopPopOptimizer, 1),
(CosineKNNOptimizer, 40),
(AsymmetricCosineKNNOptimizer, 40),
(TverskyIndexKNNOptimizer, 40),
(P3alphaOptimizer, 40),
(RP3betaOptimizer, 40),
(IALSOptimizer, 40),
(DenseSLIMOptimizer, 20),
# (SLIMOptimizer, 40), time consuming
# (MultVAEOptimizer, 5),
]
for optimizer_class, n_trials in test_configs:
recommender_name = optimizer_class.recommender_class.__name__
optimizer: BaseOptimizer = optimizer_class(data_train.X_all,
valid_evaluator)
(best_param,
validation_result_df) = optimizer.optimize(timeout=14400,
n_trials=n_trials)
validation_result_df["recommender_name"] = recommender_name
validation_results.append(validation_result_df)
pd.concat(validation_results).to_csv(f"validation_scores.csv")
test_recommender = optimizer.recommender_class(X_train_val_all,
**best_param)
test_recommender.learn()
test_scores = test_evaluator.get_scores(test_recommender, [5, 10, 20])
test_results.append(
dict(name=recommender_name, best_param=best_param, **test_scores))
with open("test_results.json", "w") as ofs:
json.dump(test_results, ofs, indent=2)