This page analyzes the hyperparameter tuning results for biased matrix factorization with ALS.
Parameter Search Space
/home/mde48/lenskit/lenskit-codex/.venv/lib/python3.12/site-packages/ray/tune/search/sample.py:700: RayDeprecationWarning: The `base` argument is deprecated. Please remove it as it is not actually needed in this method.
| embedding_size |
Integer |
LogUniform |
4 ≤ \(x\) ≤ 512 |
450 |
| regularization.user |
Float |
LogUniform |
1e-05 ≤ \(x\) ≤ 1 |
0.323 |
| regularization.item |
Float |
LogUniform |
1e-05 ≤ \(x\) ≤ 1 |
0.00317 |
| damping.user |
Float |
LogUniform |
1e-12 ≤ \(x\) ≤ 100 |
3.74 |
| damping.item |
Float |
LogUniform |
1e-12 ≤ \(x\) ≤ 100 |
1.28e-10 |
Final Result
Searching selected the following configuration:
{
'embedding_size': 450,
'regularization': {'user': 0.32339547493145027, 'item': 0.003169704945518609},
'damping': {'user': 3.7448453198454534, 'item': 1.27559519714575e-10},
'epochs': 6
}
With these metrics:
{
'RBP': 0.0006888496762251515,
'DCG': 7.726449284567924,
'NDCG': 0.265924279382336,
'RecipRank': 0.004507232900513236,
'Hit10': 0.005970149253731343,
'RMSE': 0.8055420517921448,
'max_epochs': 30,
'epoch_train_s': 180.29132019299868,
'epoch_measure_s': 343.33556148700154,
'done': True,
'training_iteration': 6,
'trial_id': '38bad571',
'date': '2025-07-28_21-14-58',
'timestamp': 1753751698,
'time_this_iter_s': 523.6316566467285,
'time_total_s': 3061.792503118515,
'pid': 132395,
'hostname': 'CCI-ws21',
'node_ip': '10.248.127.152',
'config': {
'embedding_size': 450,
'regularization': {'user': 0.32339547493145027, 'item': 0.003169704945518609},
'damping': {'user': 3.7448453198454534, 'item': 1.27559519714575e-10},
'epochs': 6
},
'time_since_restore': 3061.792503118515,
'iterations_since_restore': 6
}
Parameter Analysis
Embedding Size
The embedding size is the hyperparameter that most affects the model’s fundamental logic, so let’s look at performance as a fufnction of it:
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?
