Table 3 Barlow Twins hyperparameters

enc_n_neurons

1024, 2048, 4096

enc_n_layers

1, 2, 3

proj_n_neurons

1024, 2048, 4096

proj_n_layers

1, 2, 3

embedding_dim

512, 1024, 2048

act_function

ReLU

aa_emb_size

1024

loss_weight

1 \(\times 10^{-5}\), 0.005, 0.1

batch_size

4096

epochs

250

optimizer

AdamW

learning_rate

1 \(\times 10^{-5}\), 3 \(\times 10^{-4}\), 0.1

beta_1

0.9

beta_2

0.999

weight_decay

5\(\times 10^{-5}\)

step_size

10

gamma

0.1

val_split

0.1