ViMoE: An Empirical Study of Designing Vision Mixture-of-Experts
Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel pith:JOTBVBRLrecord.jsonopen to challenge →
read the original abstract
Mixture-of-Experts (MoE) models embody the divide-and-conquer concept and are a promising approach for increasing model capacity, demonstrating excellent scalability across multiple domains. In this paper, we integrate the MoE structure into the classic Vision Transformer (ViT), naming it ViMoE, and explore the potential of applying MoE to vision through a comprehensive study on image classification and semantic segmentation. However, we observe that the performance is sensitive to the configuration of MoE layers, making it challenging to obtain optimal results without careful design. The underlying cause is that inappropriate MoE layers lead to unreliable routing and hinder experts from effectively acquiring helpful information. To address this, we introduce a shared expert to learn and capture common knowledge, serving as an effective way to construct stable ViMoE. Furthermore, we demonstrate how to analyze expert routing behavior, revealing which MoE layers are capable of specializing in handling specific information and which are not. This provides guidance for retaining the critical layers while removing redundancies, thereby advancing ViMoE to be more efficient without sacrificing accuracy. We aspire for this work to offer new insights into the design of vision MoE models and provide valuable empirical guidance for future research.
This paper has not been read by Pith yet.
Forward citations
Cited by 7 Pith papers
-
Beyond Routing: Characterising Expert Tuning and Representation in Vision Mixture-of-Experts
Expert specialization in vision MoE models is dominated by a stable animate-inanimate distinction visible from gating to readout, with broader tuning to continuous visual and semantic dimensions rather than narrow cat...
-
FaceMoE: Mixture of Experts for Low-Resolution Face Recognition
FaceMoE introduces a MoE transformer with top-k routed specialized FFN experts for resolution-aware feature extraction in low-resolution face recognition, outperforming prior methods on eleven datasets.
-
ExPLoRe: Expert Patch-Level Loss Routing for Multi-Objective Masked Image Modeling
ExPLoRe turns MoE dispatch weights into per-patch loss coefficients for multi-objective masked image modeling, reporting gains on ImageNet-1K and ADE20K transfer.
-
Design and Behavior of Sparse Mixture-of-Experts Layers in CNN-based Semantic Segmentation
Patch-wise sparse MoE layers in CNNs for semantic segmentation yield architecture-dependent gains up to 3.9 mIoU on Cityscapes and BDD100K with low overhead, but show strong design sensitivity.
-
DoReMi: Bridging 3D Domains via Topology-Aware Domain-Representation Mixture of Experts
DoReMi uses self-supervised pre-training on topological and texture variations plus domain-aware experts with spatial-guided routing and entropy-controlled allocation to reach 80.1% mIoU on ScanNet and 77.2% mIoU on S3DIS.
-
When Does Sparse MoE Help in Vision? The Role of Backbone Compute Leverage in Sparse Routing
Sparse MoE vision models show positive accuracy gaps only when routing a substantial compute fraction ρ and using k≥2 experts at large scale; batch-axis dispatch is identified as a key failure mode.
-
Mamoda2.5: Enhancing Unified Multimodal Model with DiT-MoE
Mamoda2.5 is a 25B-parameter DiT-MoE unified AR-Diffusion model that reaches top video generation and editing benchmarks with 4-step inference up to 95.9x faster than baselines.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.