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 |
182 |
| regularization.user |
Float |
LogUniform |
1e-05 ≤ \(x\) ≤ 1 |
0.135 |
| regularization.item |
Float |
LogUniform |
1e-05 ≤ \(x\) ≤ 1 |
0.0372 |
| damping.user |
Float |
LogUniform |
1e-12 ≤ \(x\) ≤ 100 |
1.38e-06 |
| damping.item |
Float |
LogUniform |
1e-12 ≤ \(x\) ≤ 100 |
4.15 |
Final Result
Searching selected the following configuration:
{
'embedding_size': 182,
'regularization': {'user': 0.13457932006936438, 'item': 0.03716261973856335},
'damping': {'user': 1.3842050969711814e-06, 'item': 4.147629054926619},
'epochs': 6
}
With these metrics:
{
'RBP': 0.03617969967228035,
'LogRBP': 0.4749828697903258,
'NDCG': 0.30264294670384995,
'RecipRank': 0.11096809011223152,
'RMSE': 0.7983685081718532,
'TrainTask': 'fba78a57-9449-4303-af1e-a9d8a497eec1',
'TrainTime': None,
'TrainCPU': None,
'max_epochs': 30,
'done': True,
'training_iteration': 6,
'trial_id': '2d29da02',
'date': '2025-05-07_04-38-38',
'timestamp': 1746607118,
'time_this_iter_s': 272.09115052223206,
'time_total_s': 1640.7825350761414,
'pid': 157071,
'hostname': 'CCI-ws21',
'node_ip': '10.248.127.152',
'config': {
'embedding_size': 182,
'regularization': {'user': 0.13457932006936438, 'item': 0.03716261973856335},
'damping': {'user': 1.3842050969711814e-06, 'item': 4.147629054926619},
'epochs': 6
},
'time_since_restore': 1640.7825350761414,
'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?
