DCT-based initialization and frequency truncation for self-attention improve accuracy and reduce overhead in Vision Transformers on standard benchmarks.
Zero initialization: Initializing neural networks with only zeros and ones.Transactions on Machine Learning Research,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Discrete Cosine Transform Based Decorrelated Attention for Vision Transformers
DCT-based initialization and frequency truncation for self-attention improve accuracy and reduce overhead in Vision Transformers on standard benchmarks.