20 3. DATA-DRIVEN COMPATIBILITY MODELING
0.0
-0.2
-0.4
-0.6
-0.8
-1.0
1.0
0.9
0.8
0.7
0.6
0.5
Loss
AUC
Epoch Epoch
0 5 10 15 20 25 30 0 5 10 15 20 25 30
(a) Training loss. (b) AUC.
Figure 3.4: Training loss and AUC with respect to each epoch.
Knee-length skirts
Mini skirts
Hoodies
Blouses
Pants
T-shirts
Tank tops
Cardigans
Blazers
Day dresses
Cocktail dresses
Dresses
Coats
Jackets
Sweaters
Sweatshirts
Leggings
Tops
Shorts
Boyfriend jeans
Jeans
Skinny jeans
Frame jeans
Men’s jeans
Men’s coats
Men’s jackets
Figure 3.5: Illustration of the most popular matchings pairs between top and bottom categories.
3.4.2 DATA INSIGHTS
Due to the limited space, we only presented the most popular matching pairs of top and bottom
categories
2
in our dataset in Figure 3.5. Each circle denotes a fashion category, where the light
blue refers to the top categories and the dark blue denotes the bottom ones. e areas of the
2
Here we only consider the category at the finest granularity for each item.
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.