Bias Predictor

This page analyzes the behavior and hyperparameter search for the Bias scoring model on ML1M.

Parameter Search Space

Parameter	Type	Distribution	Values	Selected
damping.user	Float	LogUniform	0.1 ≤ \(x\) ≤ 100	3.93
damping.item	Float	LogUniform	0.1 ≤ \(x\) ≤ 100	2.16

Final Result

Searching selected the following configuration:

{'damping': {'user': 3.9308126900054, 'item': 2.1642801979912143}}

With these metrics:

{
    'RBP': 0.01230276377806422,
    'LogRBP': -0.6036913746690882,
    'NDCG': 0.1753524329189794,
    'RecipRank': 0.043849184373969285,
    'RMSE': 0.8574553042687229,
    'TrainTask': 'cb27dab0-cbab-40b4-9ad9-b291b57c5108',
    'TrainTime': 0.6784966590348631,
    'TrainCPU': 0.6261379999999996,
    'TestTask': '6109724d-9569-403f-88ae-2a3eacb5a4fc',
    'TestTime': 2.8886236069956794,
    'TestCPU': 2.813915,
    'timestamp': 1746436478,
    'checkpoint_dir_name': None,
    'done': True,
    'training_iteration': 1,
    'trial_id': '483a9_00085',
    'date': '2025-05-05_05-14-38',
    'time_this_iter_s': 5.163275241851807,
    'time_total_s': 5.163275241851807,
    'pid': 3341992,
    'hostname': 'CCI-ws21',
    'node_ip': '10.248.127.152',
    'config': {'damping': {'user': 3.9308126900054, 'item': 2.1642801979912143}},
    'time_since_restore': 5.163275241851807,
    'iterations_since_restore': 1,
    'experiment_tag': '85_item=2.1643,user=3.9308'
}

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	3.930813	2.16428	0.012303	0.857455

If we instead searched for RBP, we would select:

	config.damping.user	config.damping.item	RBP	RMSE
Method
Random	5.125235	96.629343	0.022169	0.867455

Search Geometry

What is the geometry of the search space?