Experiments (common)

ADDA (Ivan)

See Ivan's experiments -> ADDA

DANN - MNIST (Dima)

See Dima's experiments

DANN - AlexNet(Boris, Katya, Masha)

A->W experiments with different dropout

EXP name	domain head	adaptation block usage	domain dropout	class dropout	Source best acc	Target best acc	Source final acc	Target final acc
Alexnet_vanilla (Katya)	vanilla-dann	true	false	false	1.0	0.59375	0.9989	0.48698
Alexnet_domain_dropout (Katya)	dropout_dann	true	true	false	1.0	0.61719	1.0	0.58203
Alexnet_domain_dropout (Boris)	dropout_dann	true	true	false	?	0.61719	?	?
Alexnet_domain_dropout (Masha)	dropout_dann	true	true	false	?	?	?	0.595 (mean of several)
Paper DANN	dropout_dann	true	true	?	-	0.73	-	-
Alexnet_domain_and_class_dropout (Katya)	dropout_dann	true	true	true	1.0	0.62370	1.0	0.59375

AlexNet on different domains

	A -> W	D -> W	W -> D	A -> D	D -> A	W -> A
DANN Paper	0.73	0.964	0.992	-	-	-
Our (Boris)	0.61719	0.88281	0.95625	0.55625	0.41619	0.45668
Our (Masha)	0.595	0.94	-	-	-	-

A->W and D->W DANN and Source/Target only (Masha):

A->W

(Среднее значение итоговой метрики в конце 200 эпохи)

Результаты	DANN	Source-Only	Target-Only
Наш эксперимент	0.595	0.4518	0.95
Статья	0.73	0.642	-

D->W

(Значение итоговой метрики в конце 200 эпохи, без усреднения, только по одному запуску)

Результаты	DANN	Source-Only	Target-Only
Наш эксперимент	0.94	0.934	0.975
Статья	0.964	0.961	-

AlexNet with different bottleneck size (Masha)

Masha's experiments - ResNet50 Сравнение архитектур с размерностью в bottleneck-слое 256 (из статьи) и 2048

DANN - ResNet50(Anya, Boris, Ivan, Katya, Masha)

Common table

Model	Frozen	Bottleneck	Domain head	Class head	A → W	D→ W	W→ D	A→ D	D → A	W→ A
DANN Alexnet Статья	?	?	?	?	0.73	0.964	0.992
DANN Resnet Обзор	?	?	?	?	0.826	0.978	1	0.833	0.668	0.661
OneDomain ResNet Frozen (Anya)	all	?	-	-	0.67839	0.92318	0.98884	0.71652	0.62003	0.63388
OneDomain ResNet Frozen 129 (Anya)	129	?	-	-	0.74740	0.97005	0.99777	0.76786	0.63317	0.64453
Source-Only ResNet50 Bottleneck 256 (Masha)	?	256	-	-	0.783
Target-Only ResNet50 Bottleneck 256 (Masha)	?	256	-	-	0.987
DANN Resnet Frozen (Anya)	all	?	?	?	0.73438	0.95573	0.99554	0.77455	0.62749	0.63175
DANN Resnet Frozen 141 (Anya)	141	?	?	?	0.76172	0.97526	0.99777	0.78125	0.65270	0.63849
DANN Resnet Frozen 129 (Anya)	129	?	?	?	0.76693	0.98307	1.0	0.77679	0.65554	0.63849
DANN ResNet 0 freeze without domain loss (Ivan-exp2)	0	?	?	?	0.5443	-	-	-	-	-
DANN ResNet 72 freeze without domain loss (Ivan-exp2)	72	?	?	?	0.7348	-	-	-	-	-
DANN ResNet 129 freeze without domain loss (Ivan-exp2)	129	?	?	?	0.7849	-	-	-	-	-
DANN ResNet 141 freeze with domain loss (Ivan-exp2)	141	?	?	?	0.71484	-	-	0.62708	0.63246	0.67543
DANN ResNet 141 freeze without domain loss (Ivan-exp2)	141	?	?	?	0.77995	0.96875	0.99583	0.79583	0.63565	0.63175
DANN Rich ResNet Frozen (Anya)	all	?	?	?	0.74740
DANN Rich ResNet Frozen 141 (Anya)	141	?	?	?	0.85677					0.70916
DANN Rich ResNet Frozen 129 (Anya)	129	?	?	?	0.85807
DANN Rich ResNet Frozen 72(Anya)	72	?	?	?	0.83464
DANN Rich ResNet Frozen 141 Bottleneck 128 (Anya)	141	128	?	?	0.81771
DANN Rich ResNet Frozen 141 Bottleneck 256 (Anya)	141	256	?	?	0.86328					0.70881
DANN Rich ResNet Frozen 141 Bottleneck 512 (Anya)	141	512	?	?	0.84896					0.70810
DANN Rich ResNet Frozen 141 Bottleneck 1024 (Anya)	141	1024	?	?	0.87891					0.71626
DANN Rich ResNet Frozen 141 Bottleneck 2048 (Anya)	141	2048	?	?	0.86068					0.6985
Resnet_vanilla(Katya)	?	?	vanilla	vanilla	0.78255
Resnet_domain_dropout(Katya)	?	?	dropout	vanilla	0.85286
Resnet_domain_and_class_dropout(Katya)	?	?	dropout	dropout	0.85156
DANN ResNet50 domain dropout Bottleneck 256 (Masha)	?	256	dropout	?	0.857
dropout domain head (Boris)	?	?	dropout	?	0.82292	0.97656	0.99375	0.78125	0.64453	0.66264
ResNet with domain loss (Ivan-exp4)	?	?	?	?	0.71484			0.62708	0.63246	0.67543
ResNet without domain loss (Ivan-exp4)	?	?	?	?	0.77995	0.96875	0.99583	0.79583	0.63565	0.63175
vanilla domain head - 141 layer freeze(Ivan-exp1)	141	?	vanilla	?	0.8451	0.9349	0.9754	0.7969	0.6719	0.6893
without domain loss 141 layer freeze (Ivan-exp1)	141	?	?	?	0.7604	0.9674	0.9911	0.7924	0.6403	0.6317
test5 classificator - 141 layer freeze(Ivan-exp5)	141	?	?	?	0.8503	0.9154	0.9263	0.7902	0.6708	0.7109

A->W experiments with different dropouts (Katya)

EXP name	domain head	adaptation block usage	domain dropout	class dropout	Source best acc	Target best acc	Source final acc	Target final acc
Resnet_vanilla	vanilla-dann	false	false	false	0.99751	0.78255	0.99503	0.72786
Resnet_domain_dropout	dropout_dann	false	true	false	1.0	0.85286	1.0	0.80339
Resnet_domain_and_class_dropout	dropout_dann	false	true	true	1.0	0.85156	1.0.	0.84766

Different preprocessing for ResNet backbone (Katya)

EXP name	Source best acc	Target best acc	Source final acc	Target final acc
resnet50_no_crop	1.0	0.85156	1.0	0.84766
resnet50_with_crop	1.0	0.83984	1.0	0.82161