Understanding the Mixture-of-Experts with Nadaraya-Watson Kernel

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zheng, Chuanyang, Sun, Jiankai, Gao, Yihang, Xie, Enze, Wang, Yuehao, Wang, Peihao, Xu, Ting, Chang, Matthew, Ren, Liliang, Li, Jingyao, Xiong, Jing, Rasul, Kashif, Schwager, Mac, Schneider, Anderson, Wang, Zhangyang, Nevmyvaka, Yuriy
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911208256307200
author Zheng, Chuanyang
Sun, Jiankai
Gao, Yihang
Xie, Enze
Wang, Yuehao
Wang, Peihao
Xu, Ting
Chang, Matthew
Ren, Liliang
Li, Jingyao
Xiong, Jing
Rasul, Kashif
Schwager, Mac
Schneider, Anderson
Wang, Zhangyang
Nevmyvaka, Yuriy
author_facet Zheng, Chuanyang
Sun, Jiankai
Gao, Yihang
Xie, Enze
Wang, Yuehao
Wang, Peihao
Xu, Ting
Chang, Matthew
Ren, Liliang
Li, Jingyao
Xiong, Jing
Rasul, Kashif
Schwager, Mac
Schneider, Anderson
Wang, Zhangyang
Nevmyvaka, Yuriy
contents Mixture-of-Experts (MoE) has become a cornerstone in recent state-of-the-art large language models (LLMs). Traditionally, MoE relies on $\mathrm{Softmax}$ as the router score function to aggregate expert output, a designed choice that has persisted from the earliest MoE models to modern LLMs, and is now widely regarded as standard practice. However, the necessity of using $\mathrm{Softmax}$ to project router weights into a probability simplex remains an unchallenged assumption rather than a principled design choice. In this work, we first revisit the classical Nadaraya-Watson regression and observe that MoE shares the same mathematical formulation as Nadaraya-Watson regression. Furthermore, we show that both feed-forward neural network (FFN) and MoE can be interpreted as a special case of Nadaraya-Watson regression, where the kernel function corresponds to the input neurons of the output layer. Motivated by these insights, we propose the \textbf{zero-additional-cost} Kernel Inspired Router with Normalization (KERN), an FFN-style router function, as an alternative to $\mathrm{Softmax}$. We demonstrate that this router generalizes both $\mathrm{Sigmoid}$- and $\mathrm{Softmax}$-based routers. \textbf{Based on empirical observations and established practices in FFN implementation, we recommend the use of $\mathrm{ReLU}$ activation and $\ell_2$-normalization in $\mathrm{KERN}$ router function.} Comprehensive experiments in MoE and LLM validate the effectiveness of the proposed FFN-style router function \methodNorm.
format Preprint
id arxiv_https___arxiv_org_abs_2509_25913
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Understanding the Mixture-of-Experts with Nadaraya-Watson Kernel
Zheng, Chuanyang
Sun, Jiankai
Gao, Yihang
Xie, Enze
Wang, Yuehao
Wang, Peihao
Xu, Ting
Chang, Matthew
Ren, Liliang
Li, Jingyao
Xiong, Jing
Rasul, Kashif
Schwager, Mac
Schneider, Anderson
Wang, Zhangyang
Nevmyvaka, Yuriy
Computation and Language
Mixture-of-Experts (MoE) has become a cornerstone in recent state-of-the-art large language models (LLMs). Traditionally, MoE relies on $\mathrm{Softmax}$ as the router score function to aggregate expert output, a designed choice that has persisted from the earliest MoE models to modern LLMs, and is now widely regarded as standard practice. However, the necessity of using $\mathrm{Softmax}$ to project router weights into a probability simplex remains an unchallenged assumption rather than a principled design choice. In this work, we first revisit the classical Nadaraya-Watson regression and observe that MoE shares the same mathematical formulation as Nadaraya-Watson regression. Furthermore, we show that both feed-forward neural network (FFN) and MoE can be interpreted as a special case of Nadaraya-Watson regression, where the kernel function corresponds to the input neurons of the output layer. Motivated by these insights, we propose the \textbf{zero-additional-cost} Kernel Inspired Router with Normalization (KERN), an FFN-style router function, as an alternative to $\mathrm{Softmax}$. We demonstrate that this router generalizes both $\mathrm{Sigmoid}$- and $\mathrm{Softmax}$-based routers. \textbf{Based on empirical observations and established practices in FFN implementation, we recommend the use of $\mathrm{ReLU}$ activation and $\ell_2$-normalization in $\mathrm{KERN}$ router function.} Comprehensive experiments in MoE and LLM validate the effectiveness of the proposed FFN-style router function \methodNorm.
title Understanding the Mixture-of-Experts with Nadaraya-Watson Kernel
topic Computation and Language
url https://arxiv.org/abs/2509.25913