Hyperparameters for EASER

We tested the following hyperparameter space:

Parameter Ranges
l2_norm [loguniform, 2.72, 16]

Dataset l2_norm
MovieLens 1M 696.785
Amazon Digital Music 15.930
Epinions 174.179


Hyperparameters for MultiDAE

We tested the following hyperparameter space:

Parameter Ranges
Epoch [20, 30, 40]
Learning Rate [0.0001, 0.0005,0.003, 0.001]
Batch Size [32, 64, 128, 256, 512]
intermediate_dim 600
Latent_dim [100, 200, 300]
Reg_lambda 0.01
Dropout_pkeep 1

Dataset Epoch Learning Rate Batch Size Intermediate_dim Latent_dim Reg_lambda Dropout_pkeep
MovieLens 1M 20 0.0005 128 600 100 0.01 1
Amazon Digital Music 40 0.0005 512 600 300 0.01 1
Epinions 40 0.0005 512 600 300 0.01 1

Hyperparameters for MultiVAE


Dataset Epoch Learning Rate Batch Size Intermediate_dim Latent_dim Reg_lambda Dropout_pkeep
MovieLens 1M - - - - - - -
Amazon Digital Music - - - - - - -
Epinions - - - - - - -

Hyperparameters for GMF

We tested the following hyperparameter space:

Parameter Ranges
Epochs [30, 40, 50]
Learning Rate [0.0001, 0.0005, 0.001]
Batch Size [128, 256, 512]
Mf_factors [8, 16, 32, 64, 128, 256]
Is_edge_weight_train True

Dataset Epochs Learning Rate Batch Size Mf_factors Is_edge_weight_train
MovieLens 1M 40 512 128 0.00014 true
Amazon Digital Music 50 256 256 0.00054 true
Epinions 40 512 256 0.00035 True

Hyperparameters for NeuMF

We tested the following hyperparameter space:

Parameter Ranges
Epoch [quniform, 30, 100, 1]
Learning rate [loguniform, -11.512925464970229, 0]
Mf_factors [64, 128, 256]
Batch Size [64, 128, 256]
Dropout 0
Is_mf_train True
Is_mlp_train True
m [4,6,8]


Dataset Epoch Learning rate Mf_factors Batch Size Dropout Is_mf_train Is_mlp_train m
MovieLens 1M - - - - - - - -
Amazon Digital Music - - - - - - - -
Epinions - - - - - - - -

Hyperparameters for ConvNeuMF

We tested the following hyperparameter space:

Parameter Ranges
Epoch [20, 30]
Learning Rate [0.0001, 0.0005, 0.001]
Batch Size [16, 32]
Embedding Size [16, 32]
l_w [0.0001, 0.001, 0.0005]
l_b [0.01, 0.001,0.0001, 0.00001]
Dropout [0.1,0.2,0.3, 0.4, 0.5]

Dataset Epoch Learning Rate Batch Size Embedding Size l_w l_b Dropout
MovieLens 1M 30 0.0005 16 32 0.0001 0.01 0.7
Amazon Digital Music 30 0.0005 16 32- 0.0005 0.001 0.5
Epinions 25 0.0001 16 100 0.1 0.001 0.45

Hyperparameters for ConvMF

Parameter Ranges
Epoch [20, 30]
Learning Rate [0.0001, 0.0005, 0.001, 0.003]
Batch Size [50, 64, 100]
Embedding Size [64, 128, 256]
l_w [0.0001, 0.001, 0.0005]
l_b [0.01, 0.001,0.0001, 0.00001]
Dropout [0.1,0.2,0.3, 0.4, 0.5]

Dataset Epoch Learning Rate Batch Size Embedding Size l_w l_b Dropout
MovieLens 1M 25 0.0001 32 100 0.001 0.00001 0.2
Amazon Digital Music 25 0.0001 16 64 0.001 0.00001 0.2
Epinions 25 0.0001 16 100 0.1 0.001 0.45

Hyperparameters for NGCF

We tested the following hyperparameter space:

Parameter Options
Learning rate [0.0001, 0.0005, 0.001, 0.003]
Epoch [30, 40, 50]
Batch Size [64, 128, 256, 512]
Factors [8, 16, 32, 64, 128, 256]
Node_dropout [0.1,0.2, 0.3, 0.45, 0.5]
Message_dropout [0.1,0.2, 0.3, 0.45, 0.5]

Dataset Epoch Learning Rate Batch Size Factors
MovieLens 1M 30 0.0002 512 64
Amazon Digital Music 40 0.00005 256 32
Epinions 50 0.0002 512 8