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 |
330 |
| regularization.user |
Float |
LogUniform |
1e-05 ≤ \(x\) ≤ 1 |
0.0903 |
| regularization.item |
Float |
LogUniform |
1e-05 ≤ \(x\) ≤ 1 |
0.0392 |
| damping.user |
Float |
LogUniform |
1e-12 ≤ \(x\) ≤ 100 |
1.88e-05 |
| damping.item |
Float |
LogUniform |
1e-12 ≤ \(x\) ≤ 100 |
0.000165 |
Final Result
Searching selected the following configuration:
{
'embedding_size': 330,
'regularization': {'user': 0.09026204464706182, 'item': 0.039227647579904934},
'damping': {'user': 1.8785553295225284e-05, 'item': 0.00016458494355190166},
'epochs': 6
}
With these metrics:
{
'RBP': 0.0016840378949438514,
'LogRBP': -2.592320890707648,
'NDCG': 0.1615156871747588,
'RecipRank': 0.01622092087443804,
'RMSE': 0.8121416182411427,
'TrainTask': '3f3b4b9b-fbb3-4590-8c30-ba30a5553e0d',
'TrainTime': None,
'TrainCPU': None,
'max_epochs': 30,
'done': True,
'training_iteration': 6,
'trial_id': '4c049cef',
'date': '2025-05-05_00-32-43',
'timestamp': 1746419563,
'time_this_iter_s': 4.655622959136963,
'time_total_s': 26.12876296043396,
'pid': 2797815,
'hostname': 'CCI-ws21',
'node_ip': '10.248.127.152',
'config': {
'embedding_size': 330,
'regularization': {'user': 0.09026204464706182, 'item': 0.039227647579904934},
'damping': {'user': 1.8785553295225284e-05, 'item': 0.00016458494355190166},
'epochs': 6
},
'time_since_restore': 26.12876296043396,
'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?
