FlexMF Logistic on ML1M

This page analyzes the hyperparameter tuning results for the FlexMF scorer in implicit-feedback mode with logistic loss (Logistic Matrix Factorization).

Parameter Search Space

Parameter	Type	Distribution	Values	Selected
embedding_size	Integer	LogUniform	4 ≤ \(x\) ≤ 512	109
regularization	Float	LogUniform	0.0001 ≤ \(x\) ≤ 10	0.405
learning_rate	Float	LogUniform	0.001 ≤ \(x\) ≤ 0.1	0.00832
reg_method	Categorical	Uniform	L2, AdamW	AdamW
negative_count	Integer	Uniform	1 ≤ \(x\) ≤ 5	4
positive_weight	Float	Uniform	1 ≤ \(x\) ≤ 10	4.05
user_bias	Categorical	Uniform	True, False	True
item_bias	Categorical	Uniform	True, False	False

Final Result

Searching selected the following configuration:

{
    'embedding_size': 109,
    'regularization': 0.405361839333611,
    'learning_rate': 0.008321602556951814,
    'reg_method': 'AdamW',
    'negative_count': 4,
    'positive_weight': 4.048579386127648,
    'user_bias': True,
    'item_bias': False,
    'epochs': 26
}

With these metrics:

{
    'RBP': 0.10479121677137343,
    'NDCG': 0.35969290195214815,
    'RecipRank': 0.35549524449671305,
    'TrainTask': '1dcb5145-0753-458b-8733-38f5be0274d1',
    'TrainTime': None,
    'TrainCPU': None,
    'max_epochs': 50,
    'done': False,
    'training_iteration': 26,
    'trial_id': 'c9efa756',
    'date': '2025-04-03_21-47-32',
    'timestamp': 1743731252,
    'time_this_iter_s': 1.3986763954162598,
    'time_total_s': 42.324442625045776,
    'pid': 1202393,
    'hostname': 'CCI-ws21',
    'node_ip': '10.248.127.152',
    'config': {
        'embedding_size': 109,
        'regularization': 0.405361839333611,
        'learning_rate': 0.008321602556951814,
        'reg_method': 'AdamW',
        'negative_count': 4,
        'positive_weight': 4.048579386127648,
        'user_bias': True,
        'item_bias': False,
        'epochs': 26
    },
    'time_since_restore': 5.9172282218933105,
    'iterations_since_restore': 4
}