ダイナミックニューラルネットワーク
2024-11-29 09:48:24 71 0 報告 0
0
ログインして完全な内容を表示
著者の他の作品
概要/内容
How to be dynamic
Dynamic Architecture
Cascading
Early exiting
Layer skipping
Dynamic prunning
Dynamic parameter
Convolution kernel
Linear transformation
Activation function
Dynamic input processing
Image input
Sequence input
Training
Reinforcement learning
Gumbel-softmax
Improved SemHash
Knowledge distillation
Application
CV
Image Segmention
Object detection
Image synthesis
Video recognition
Action detection and action spotting
Others
NLP
Text classfication
Question answering
Large languag model
Sentiment analysis
Others
Challenges
Huge storage space
Lighter-weight
Architecture Design
co-design with hardware
Stability
Combining with popular models
集める
集める
集める
集める
0 コメント
次のページ
