Bias Predictor

This page analyzes the behavior and hyperparameter search for the Bias scoring model on ML32M.

Parameter Search Space

Parameter	Type	Distribution	Values	Selected
damping.user	Float	LogUniform	0.1 ≤ \(x\) ≤ 100	5.82
damping.item	Float	LogUniform	0.1 ≤ \(x\) ≤ 100	9.66

Final Result

Searching selected the following configuration:

{'damping': {'user': 5.818752728872168, 'item': 9.658111442890217}}

With these metrics:

{
    'RBP': 0.07449020357740353,
    'LogRBP': 1.197152311923512,
    'NDCG': 0.3236420702720456,
    'RecipRank': 0.21058885299020655,
    'RMSE': 0.8285240417086436,
    'TrainTask': '4066226b-e375-41ed-9227-725f49450d1e',
    'TrainTime': 5.3903352429988445,
    'TrainCPU': 5.395483,
    'TestTask': 'c993615e-8d3f-4b57-9629-2ab42ea68dc4',
    'TestTime': 27.41916402700008,
    'TestCPU': 27.495500000000003,
    'timestamp': 1746608284,
    'checkpoint_dir_name': None,
    'done': True,
    'training_iteration': 1,
    'trial_id': '5697d7c6',
    'date': '2025-05-07_04-58-04',
    'time_this_iter_s': 62.27134680747986,
    'time_total_s': 62.27134680747986,
    'pid': 174052,
    'hostname': 'CCI-ws21',
    'node_ip': '10.248.127.152',
    'config': {'damping': {'user': 5.818752728872168, 'item': 9.658111442890217}},
    'time_since_restore': 62.27134680747986,
    'iterations_since_restore': 1,
    'experiment_tag': '15_item=9.6581,user=5.8188'
}

Metric Response

How does RMSE change with each setting independently?

User Damping

Item Damping

Best Configurations

Since this is an explicit-feedback rating prediction model, our primary search criteria is RMSE. The configuration with the best RMSE is:

	config.damping.user	config.damping.item	RBP	RMSE
Method
Random	5.818753	9.658111	0.07449	0.828524

If we instead searched for RBP, we would select:

	config.damping.user	config.damping.item	RBP	RMSE
Method
Random	0.527982	49.674723	0.081341	0.832032

Search Geometry

What is the geometry of the search space?