2022 · 23 citations · 26 references
ComputingComputer ScienceArtificial IntelligenceMachine LearningNatural Language ProcessingAttention NetworksDeep LearningLanguage ModelingLanguage ModelsMachine TranslationNeural NetworksAttentionAttention MechanismsComputational EfficiencyMasked Language ModelingModel CapacityNeural Language ModelsTransformer Language ModelsTransformer Models
Mixture-of-Experts (MoE) networks have been proposed as an efficient way to scale up model capacity and implement conditional computing. However, the study of MoE components mostly focused on the feedforward layer in Transformer architecture. This paper proposes the Mixture of Attention Heads (MoA), a new architecture that combines multi-head attention with the MoE mechanism. MoA includes a set of attention heads that each has its own set of parameters. Given an input, a router dynamically selects a subset of k attention heads per token. This conditional computation schema allows MoA to achieve stronger performance than the standard multi-head attention layer. Furthermore, the sparsely gated MoA can easily scale up the number of attention heads and the number of parameters while preserving computational efficiency. Despite performance improvements, MoA also automatically differentiates heads' utilities, providing a new perspective to discuss the model's interpretability. We conducted experiments on several important tasks, including Machine Translation and Masked Language Modeling. Experiments have shown promising results on several tasks against strong baselines that involve large and very deep models.
26
Sepp Hochreiter, Jürgen Schmidhuber · Neural Computation · 1997 · 98.1K citations
Adam: A Method for Stochastic Optimization
Diederik P. Kingma, Jimmy Ba · UvA-DARE (University of Amsterdam) · 2014 · 84.8K citations · Full text
DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · 2023 · 76.3K citations · Full text
Rethinking the Inception Architecture for Computer Vision
Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe et al. · 2016 · 30.9K citations
Kishore Papineni, Salim Roukos, Todd J. Ward et al. · 2001 · 21.5K citations · Full text