Regeneration Learning
Regeneration Learning: A Learning Paradigm for Data Generation
Transformer Architecture
ResiDual: Transformer with Dual Residual Connections
A Study on ReLU and Softmax in Transformer
Transformer as Memories
Transformer: A Study on ReLU and Softmax in Transformer
Speech Synthesis: NaturalSpeech (MemoryVAE)
Speech Synthesis: AdaSpeech 4 (Speaker Memory)
Text Generation: nkNN-NMT (n-gram kNN Memory)
Text Generation: Extract and Attend (Dictionary as Memory)
Talking-Face Video Synthesis: MemFace (Memories as One-to-Many Alleviators)
Diffusion Models
Speedup: PriorGrad, InferGrad, ResGrad
Speedup: ERA-Solver
