experiments.txt

Experiment id's with their corresponding parameter values (when different from default):
1 - unet target_type=negative, sgd optimizer
2 - gan 200 epochs, adam optimizer
3 - gan 200 epochs, l2 edge loss (penalty=100)
4 - gan 200 epochs, l2 edge loss (penalty=15)
5 - gan 200 epochs, no edge loss, more dropout
6 - gan 200 epochs, l2 edge loss (penalty=15), more dropout
7 - gan 20 epochs (test for lists)

(validation lists are obsolete)
8 - gan 200 epochs, no l2 loss, no const loss, no dropout
9 - gan 200 epochs, no l2 loss, const loss 15, no dropout
10 - gan 200 epochs, l2 loss 15, no const loss, no dropout
11 - gan 200 epochs, l2 loss 15, const loss 15, no dropout

(validation lists are obsolete)
12 - gan 200 epochs, no l2 loss, no const loss, no dropout
13 - unet 200 epochs, no l2 loss, no const loss, no dropout
14 - unet 200 epochs, no l2 loss, no const loss, no dropout, data augmentation
15 - gan 200 epochs, no l2 loss, no const loss, no dropout, data augmentation

(Final runs)
16 - unet 200 epochs, no l2 loss, no const loss, no dropout
17 - gan 200 epochs, no l2 loss, no const loss, no dropout
18 - unet 200 epochs, l2 loss 50, no const loss, no dropout
19 - gan 200 epochs, no l2 loss, no const loss, no dropout, data augmentation
20 - unet 200 epochs, no l2 loss, no const loss, no dropout, data augmentation
21 - unet 200 epochs, no l2 loss, const loss 15, no dropout
22 - unet 200 epochs, l2 loss 15, const loss 15, no dropout

24 - unet, 667a, precise, 100 epochs, 64 filters
25 - unet, 667a, rough, 100 epochs, 64 filters
26 - unet, 667a, precise, 100 epochs, 32 filters, dropout
27 - unet, 667a, rough, 100 epochs, 32 filters, dropout
28 - gan, 667a, precise, 100 epochs, 64 filters, dropout
29 - gan, 667a, rough, 100 epochs, 64 filters, dropout
30 - gan, 667a, precise, 200 epochs, 64 filters
31 - gan, 667a, rough, 200 epochs, 64 filters

32 - special chars in val, specs of 21
33 - special chars in train, specs of 21

34 - songit styles, specs of 21

35 - special chars in train, specs of 21