ALS ImplicitMF on ML100K

This page analyzes the hyperparameter tuning results for implicit-feedback matrix factorization with ALS.

Parameter Search Space

Parameter	Type	Distribution	Values
embedding_size	Integer	LogUniform	4 ≤ \(x\) ≤ 512
regularization.user	Float	LogUniform	1e-05 ≤ \(x\) ≤ 1
regularization.item	Float	LogUniform	1e-05 ≤ \(x\) ≤ 1
damping.user	Float	LogUniform	1e-12 ≤ \(x\) ≤ 100
damping.item	Float	LogUniform	1e-12 ≤ \(x\) ≤ 100

Final Result

Searching selected the following configuration:

{
    'embedding_size': 14,
    'regularization': {'user': 0.00019394651737243048, 'item': 0.046443125874557636},
    'damping': {'user': 8.783956209343206e-07, 'item': 1.0653855479252574e-08},
    'weight': 5.593088639439269,
    'epochs': 7
}

With these metrics:

{
    'RBP': 0.14137430045211807,
    'LogRBP': 1.837895677041204,
    'NDCG': 0.4231426515703964,
    'RecipRank': 0.4534279168146115,
    'TrainTask': 'e80c86d7-df04-49ee-8852-b2885a06b30b',
    'TrainTime': None,
    'TrainCPU': None,
    'max_epochs': 30,
    'done': False,
    'training_iteration': 7,
    'trial_id': '25039f59',
    'date': '2025-05-06_23-46-29',
    'timestamp': 1746589589,
    'time_this_iter_s': 0.22391152381896973,
    'time_total_s': 1.7566862106323242,
    'pid': 1061166,
    'hostname': 'gracehopper1',
    'node_ip': '192.168.225.60',
    'config': {
        'embedding_size': 14,
        'regularization': {'user': 0.00019394651737243048, 'item': 0.046443125874557636},
        'damping': {'user': 8.783956209343206e-07, 'item': 1.0653855479252574e-08},
        'weight': 5.593088639439269,
        'epochs': 7
    },
    'time_since_restore': 1.7566862106323242,
    'iterations_since_restore': 7
}

Iteration Completion

How many iterations, on average, did we complete?

How did the metric progress in the best result?

How did the metric progress in the longest results?

Optimization Performance

We want to assess how quickly the models optimize, to get a sense of how many training attempts we need for larger data sets.