[Retracted] Relieving the Incompatibility of Network Representation and Classification for Long-Tailed Data Distribution

Table 2

Ablation of using different experts while applying the proposed method.

Model	Many-shot	Medium-shot	Few-shot	Acc
ResNet-10	>100	≤100 and >20	≤20
Plain model	56.8	25.7	3.6	34.6
Many-shot model	57.9	—	—	—
Medium-shot model	—	32.5	—	—
Low-shot model	—	—	10.6	—
OLTR [10]	43.2	35.1	18.5	35.6
Ours with many-shot/OLTR/OLTR	54.0	34.1	17.4	39.2
Ours with many-shot/medium-shot/low-shot	54.4	28.7	8.9	36.9
Average over combination of designed experts	53.7	32.9	14.5	37.2
Average over randomly splitting strategy	48.3	27.6	8.7	36.4

Ours with A/B/C refers to A, B, and C which are used as expert models to supervise many-shot/medium-shot/few-shot subsets, respectively. Experiments are performed on ImageNet-LT with ResNet-10.