This page analyzes the hyperparameter tuning results for the implicit-feedback ALS matrix factorization model.
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 |
29 |
| regularization.user |
Float |
LogUniform |
1e-05 ≤ \(x\) ≤ 1 |
0.000426 |
| regularization.item |
Float |
LogUniform |
1e-05 ≤ \(x\) ≤ 1 |
0.000669 |
| damping.user |
Float |
LogUniform |
1e-12 ≤ \(x\) ≤ 100 |
0.00059 |
| damping.item |
Float |
LogUniform |
1e-12 ≤ \(x\) ≤ 100 |
4.75e-11 |
| weight |
Float |
Uniform |
5 ≤ \(x\) ≤ 100 |
5.1 |
Final Result
Searching selected the following configuration:
{
'embedding_size': 29,
'regularization': {'user': 0.0004256186789587035, 'item': 0.0006685413608869978},
'damping': {'user': 0.0005900842648067328, 'item': 4.748990451426033e-11},
'weight': 5.099171622868312,
'epochs': 4
}
With these metrics:
{
'RBP': 0.19025128426592744,
'LogRBP': 2.1348304379208702,
'NDCG': 0.441576760655177,
'RecipRank': 0.36493155503425506,
'TrainTask': '0df97675-64d0-4dd1-ac5c-16311ab09df0',
'TrainTime': None,
'TrainCPU': None,
'max_epochs': 30,
'done': False,
'training_iteration': 4,
'trial_id': '269b1cd6',
'date': '2025-05-05_22-57-44',
'timestamp': 1746500264,
'time_this_iter_s': 32.28228139877319,
'time_total_s': 145.5365447998047,
'pid': 407079,
'hostname': 'CCI-ws21',
'node_ip': '10.248.127.152',
'config': {
'embedding_size': 29,
'regularization': {'user': 0.0004256186789587035, 'item': 0.0006685413608869978},
'damping': {'user': 0.0005900842648067328, 'item': 4.748990451426033e-11},
'weight': 5.099171622868312,
'epochs': 4
},
'time_since_restore': 145.5365447998047,
'iterations_since_restore': 4
}
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?
