FlexMF WARP

This page analyzes the hyperparameter tuning results for the FlexMF scorer in implicit-feedback mode with WARP loss.

Parameter Search Space

Parameter	Type	Distribution	Values	Selected
embedding_size_exp	Integer	Uniform	3 ≤ \(x\) ≤ 10	4
regularization	Float	LogUniform	0.0001 ≤ \(x\) ≤ 10	0.117
learning_rate	Float	LogUniform	0.001 ≤ \(x\) ≤ 0.1	0.0318
reg_method	Categorical	Uniform	L2, AdamW	AdamW
item_bias	Categorical	Uniform	True, False	False

Final Result

Searching selected the following configuration:

{
    'embedding_size_exp': 4,
    'regularization': 0.11661573270761147,
    'learning_rate': 0.03181517913263124,
    'reg_method': 'AdamW',
    'item_bias': False,
    'epochs': 11
}

With these metrics:

{
    'RBP': 0.08246482905700545,
    'DCG': 1.1407944788122368,
    'NDCG': 0.3203033628625242,
    'RecipRank': 0.290380303426754,
    'Hit10': 0.5612582781456954,
    'max_epochs': 50,
    'epoch_train_s': 1.9850644499529153,
    'epoch_measure_s': 2.0529073840007186,
    'done': False,
    'training_iteration': 11,
    'trial_id': '9119d039',
    'date': '2025-09-30_12-55-40',
    'timestamp': 1759251340,
    'time_this_iter_s': 4.041879177093506,
    'time_total_s': 42.763832092285156,
    'pid': 4131656,
    'hostname': 'CCI-ws21',
    'node_ip': '10.248.127.152',
    'config': {
        'embedding_size_exp': 4,
        'regularization': 0.11661573270761147,
        'learning_rate': 0.03181517913263124,
        'reg_method': 'AdamW',
        'item_bias': False,
        'epochs': 11
    },
    'time_since_restore': 42.763832092285156,
    'iterations_since_restore': 11
}

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?