Hyperparameters for gru4rec (fixed layer size of 100)

We tested the following hyperparameter space:

Parameter From To Steps
Loss Function BPR-MAX TOP1-MAX -
Final Activation Function ELU-0.5 Linear -
Learning Rate 0.1
0.5		 0.01
0.1		 10
5
Momentum 0.00 0.90 0.10
Drop-Out 0.00 0.90 0.10
Constrained Embedding True False -

Dataset Loss Function Final Activation Function Learning Rate Momentum Drop-Out Constrained Embedding
RSC15 TOP1-MAX Linear 0.04 0.0 0.3 True
RETAILROCKET TOP1-MAX Linear 0.03 0.2 0.3 True
ZALANDO BPR-MAX ELU-0.5 0.1 0.3 0.1 False
DIGINETICA TOP1-MAX Linear 0.05 0.0 0.4 True
DIGINETICA (STAMP) TOP1-MAX ELU-0.5 0.07 0.0 0.6 True
8TRACKS TOP1-MAX ELU-0.5 0.04 0.3 0.8 True
AOTM TOP1-MAX ELU-0.5 0.04 0.6 0.0 False
NOWPLAYING TOP1-MAX ELU-0.5 0.05 0.1 0.6 True
30MUSIC TOP1-MAX ELU-0.5 0.05 0.1 0.6 True

Hyperparameters for stamp (fixed layer size of 100)

We tested the following hyperparameter space:

Parameter From To Steps
Number of Epochs 10 30 10
Decay Rate 0.0 0.9 10
Initial Learning Rate 0.001
0.0001		 0.01
0.001		 10
10

Dataset Number of Epochs Decay Rate Initial Learning Rate
RSC15 20 0 0.0007
RETAILROCKET 10 0.6 0.0008
ZALANDO 30 0.7 0.009
DIGINETICA 20 0.1 0.0009
DIGINETICA (STAMP) 20 0.1 0.0008
8TRACKS 30 0.0 0.0008
AOTM 30 0 0.004
NOWPLAYING 20 0.9 0.0005
30MUSIC 10 0.4 0.003

Hyperparameters for narm (fixed factors' number of 100, layer size of of 100, and epochs' number of 20)

We tested the following hyperparameter space:

Parameter From To Steps
Learning Rate 0.1
0.5		 0.01
0.1		 10
5

Dataset Learning Rate
RSC15 0.0008
RETAILROCKET 0.01
ZALANDO 0.007
DIGINETICA 0.0007
DIGINETICA (STAMP) 0.008
8TRACKS 0.002
AOTM 0.004
NOWPLAYING 0.004
30MUSIC 0.007

Hyperparameters for nextitnet

We tested the following hyperparameter space:

Parameter From To Steps
Learning Rate 0.01
0.001		 0.001
0.0001		 10
5
Iterations 10 30 10
Negative Sampling True False -

Dataset Learning Rate Iterations Negative Sampling
RETAILROCKET 0.006 10 True
DIGINETICA 0.003 10 False
DIGINETICA (STAMP) 0.009 20 False
AOTM 0.005 30 True

Hyperparameters for csrm (fixed embedding dimension of 100 in 10 epochs)

We tested the following hyperparameter space:

Parameter From To Steps Options
Learning Rate 0.001
0.0001		 0.0001
0.00001		 10
10		 -
Memory Size - - - 128,256,512

Dataset Learning Rate Memory Size
RSC15 0.0002 256
RETAILROCKET 0.0003 512
ZALANDO 0.0005 256
DIGINETICA 0.0002 256
8TRACKS 0.0008 256
AOTM 0.00001 256
NOWPLAYING 0.0005 128
30MUSIC 0.0009 128

Hyperparameters for sr-gnn (fixed hidden layer size of 100)

We tested the following hyperparameter space:

Parameter From To Steps
Learning Rate 0.01 0.0001 20
L2 Regularization 0.0001 0.000001 20
LR Decay 0.0 0.9 10
LR Decay Steps 3 7 3

Dataset Learning Rate L2 Regularization LR Decay LR Decay Steps
RSC15 0.0007 0.00001 0.1 7
RETAILROCKET 0.0002 0.000003 0.54 3
ZALANDO 0.006 0.000005 0.28 3
DIGINETICA 0.0001 0.000007 0.63 3
8TRACKS 0.002 0.00005 0.46 7
AOTM 0.001 0.00006 0.1 7
NOWPLAYING 0.006 0.000007 0.1 3
30MUSIC 0.006 0.00003 0.36 3

Hyperparameters for sr

We tested the following hyperparameter space:

Parameter From To Steps Options
Steps 2
14		 15
30		 13
4
Weighting - - - Div, Linear, Quadratic, Log

Dataset Steps Weighting
RSC15 8 Div
RETAILROCKET 7 Div
ZALANDO 3 Quadratic
DIGINETICA 25 Div
DIGINETICA (STAMP) 30 Quadratic
8TRACKS 25 Log
AOTM 6 Div
NOWPLAYING 9 Quadratic
30MUSIC 30 Quadratic

Hyperparameters for sknn (fixed sampling mode of recent)

We tested the following hyperparameter space:

Parameter Options
Number of Neighbors 50, 100, 500, 1000, 1500
Sample Size 500, 1000, 2500, 5000, 10000
Similarity Cosine, Jaccard

Dataset Number of Neighbors Sample Size Similarity
RSC15 500 10000 Jaccard
RETAILROCKET 50 5000 Cosine
ZALANDO 50 10000 Cosine
DIGINETICA 100 500 Cosine
DIGINETICA (STAMP) 1000 5000 Cosine
8TRACKS 1000 1000 Cosine
AOTM 50 1000 Cosine
NOWPLAYING 50 2500 Jaccard
30MUSIC 100 500 Cosine

Hyperparameters for vsknn (fixed sampling mode of recent)

We tested the following hyperparameter space:

Parameter Options
Number of Neighbors 50, 100, 500, 1000, 1500
Sample Size 500, 1000, 2500, 5000, 10000
Weighting Same, Div, Linear, Quadratic, Log
Weighting Score Same, Div, Linear, Quadratic, Log
IDF Weighting False, 1, 2, 5, 10

Dataset Number of Neighbors Sample Size Weighting Weighting Score IDF_Weighting
RSC15 100 1000 Quadratic Quadratic False
RETAILROCKET 1500 2500 Same Linear 10
ZALANDO 50 10000 Log Quadratic 10
DIGINETICA 500 5000 Quadratic Div 5
DIGINETICA (STAMP) 50 2500 Log Quadratic 1
8TRACKS 100 5000 Quaratic Quadratic False
AOTM 50 100 Div Quadratic False
NOWPLAYING 100 2500 Quadratic Quadratic False
30MUSIC 100 10000 Quadratic Quadratic False

Hyperparameters for vstan (fixed sampling mode of recent)

We tested the following hyperparameter space:

Parameter Options
Number of Neighbors 100, 200, 500, 1000, 1500, 2000
Sample Size 1000, 2500, 5000, 10000
Similarity Cosine, Dot Product
Session Position Weighting 0.00001, L/8, L/4, L/2, L, 2L (L=average session length in the dataset)
Session Neighborhood Weighting 2.5, 5, 10, 20, 40, 80, 100
Item Neighborhood Weighting 0.00001, L/8, L/4, L/2, L, 2L (L=average session length in the dataset)
Item Position Weighting 0.00001, L/8, L/4, L/2, L, 2L (L=average session length in the dataset)
IDF Weighting False, 1, 2, 5, 10

Dataset Number of Neighbors Sample Size Similarity Session Position Weighting Session Neighborhood Weighting Item Neighborhood Weighting Item Position Weighting IDF Weighting
RSC15 1500 10000 Cosine 0.5 5 2 2 2
RETAILROCKET 1000 2500 Dot Product 3.62 100 0.4525 3.62 1
ZALANDO 1500 10000 Cosine 3.13 100 3.13 1.56 1
DIGINETICA 1500 5000 Cosine 4.9 40 4.9 1.225 10
8TRACKS 2000 2500 Cosine 5.68 100 22.72 0.00001 False
AOTM 200 5000 Cosine 7.05 80 14.1 0.00001 5
NOWPLAYING 200 1000 Dot Product 10.2 40 2 1.275 False
30MUSIC 2000 2000 Dot Product 8.4 40 4.2 0.00001 1

Hyperparameters for stan (fixed sampling mode of recent)

We tested the following hyperparameter space:

Parameter Options
Number of Neighbors 100, 200, 500, 1000, 1500, 2000
Sample Size 1000, 2500, 5000, 10000
Similarity Cosine, Dot Product
Session Position Weighting 0.00001, L/8, L/4, L/2, L, 2L (L=average session length in the dataset)
Session Neighborhood Weighting 2.5, 5, 10, 20, 40, 80, 100
Item Neighborhood Weighting 0.00001, L/8, L/4, L/2, L, 2L (L=average session length in the dataset)

Dataset Number of Neighbors Sample Size Session Position Weighting Session Neighborhood Weighting Item Neighborhood Weighting
RSC15 1000 10000 0.00001 10 2
RETAILROCKET 500 1000 1.81 100 0.4525
ZALANDO 100 1000 1.56 100 3.13
DIGINETICA 500 10000 1.225 20 4.9
8TRACKS 500 10000 5.68 100 11.36
AOTM 500 1000 28.2 100 14.1
NOWPLAYING 2000 2500 0.00001 100 20.4
30MUSIC 1000 10000 0.00001 100 4.2

Hyperparameters for vsknn

We tested the following hyperparameter space:

Parameter Options
Expert StdExpert, DirichletExpert
Max Considered Context Length 5,10,20,30,40,50,75
Number of Recent Candidates (Only for Adaptive Configuration) 5,10,20,30,40,50,75

Dataset Expert Max Considered Context Length Number of Recent Candidates
All StdExpert 50 1000