HyperAI超神経

プラットフォーム

Incremental Learning On Imagenet100 10 Steps

評価指標

# M Params

Average Incremental Accuracy

Average Incremental Accuracy Top-5

Final Accuracy

Final Accuracy Top-5

評価結果

このベンチマークにおける各モデルのパフォーマンス結果

						Paper Title
TCIL	116.54	77.66	94.17	67.34	88.84	Resolving Task Confusion in Dynamic Expansion Architectures for Class Incremental Learning
TCIL-Lite	26.36	77.50	93.60	67.30	87.94	Resolving Task Confusion in Dynamic Expansion Architectures for Class Incremental Learning
DER w/o Pruning	112.27	77.18	93.23	66.70	87.52	DER: Dynamically Expandable Representation for Class Incremental Learning
DER	-	76.12	92.79	66.07	88.38	DER: Dynamically Expandable Representation for Class Incremental Learning
DyTox	11.01	77.15	92.04	69.10	87.98	DyTox: Transformers for Continual Learning with DYnamic TOken eXpansion
WA	11.22	-	91.00	-	84.10	Maintaining Discrimination and Fairness in Class Incremental Learning
BiC	11.22	-	90.60	-	84.40	Large Scale Incremental Learning
E2E	11.22	-	89.92	-	80.29	End-to-End Incremental Learning
RPSNet	-	-	87.90	-	74.00	An Adaptive Random Path Selection Approach for Incremental Learning
iCaRL	11.22	-	83.60	-	63.80	iCaRL: Incremental Classifier and Representation Learning
FOSTER	-	77.75	-	-	-	FOSTER: Feature Boosting and Compression for Class-Incremental Learning
kNN-CLIP	-	85.1	-	-	-	Revisiting a kNN-based Image Classification System with High-capacity Storage
RMM (ResNet-18)	-	78.47	-	-	-	RMM: Reinforced Memory Management for Class-Incremental Learning

0 of 13 row(s) selected.

Incremental Learning On Imagenet100 10 Steps | SOTA | HyperAI超神経