# Beyond Missing Rates: Rethinking Incomplete Multi-View Clustering with Protocol Divergence

arXiv: https://arxiv.org/abs/2606.04857 · HTML: https://mlsanity.com/paper/2606.04857/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 21 of its 40 references that are in the map.
- First posted on arXiv 3 Jun 2026.

## Builds on (most cited first, 21 of 21)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Masked Autoencoders Are Scalable Vision Learners](https://mlsanity.com/paper/2111.06377/) (2021)
- [Exploring Simple Siamese Representation Learning](https://mlsanity.com/paper/2011.10566/) (2020)
- [Deep Sets](https://mlsanity.com/paper/1703.06114/) (2017)
- [Perceiver: General Perception with Iterative Attention](https://mlsanity.com/paper/2103.03206/) (2021)
- [Are Multimodal Transformers Robust to Missing Modality?](https://mlsanity.com/paper/2204.05454/) (2022)
- [Large-scale Multi-view Subspace Clustering in Linear Time](https://mlsanity.com/paper/1911.09290/) (2019)
- [Doubly Aligned Incomplete Multi-view Clustering](https://mlsanity.com/paper/1903.02785/) (2019)
- [A Survey on Incomplete Multi-view Clustering](https://mlsanity.com/paper/2208.08040/) (2022)
- [Structured Graph Learning for Scalable Subspace Clustering: From Single-view to Multi-view](https://mlsanity.com/paper/2102.07943/) (2021)
- [Generative Partial Multi-View Clustering](https://mlsanity.com/paper/2003.13088/) (2020)
- [Spectral Perturbation Meets Incomplete Multi-view Data](https://mlsanity.com/paper/1906.00098/) (2019)
- [Incomplete Multi-view Clustering via Prototype-based Imputation](https://mlsanity.com/paper/2301.11045/) (2023)
- [CDC: A Simple Framework for Complex Data Clustering](https://mlsanity.com/paper/2403.03670/) (2024)
- [Incomplete Multi-view Clustering via Diffusion Contrastive Generation](https://mlsanity.com/paper/2503.09185/) (2025)
- [High-order Multi-view Clustering for Generic Data](https://mlsanity.com/paper/2209.10838/) (2022)
- [CDIMC-net: Cognitive Deep Incomplete Multi-view Clustering Network](https://mlsanity.com/paper/2403.19514/) (2024)
- [Imputation-free and Alignment-free: Incomplete Multi-view Clustering Driven by Consensus Semantic Learning](https://mlsanity.com/paper/2505.11182/) (2025)
- [Provable Filter for Real-world Graph Clustering](https://mlsanity.com/paper/2403.03666/) (2024)
- [Incomplete Multi-view Clustering via Hierarchical Semantic Alignment and Cooperative Completion](https://mlsanity.com/paper/2510.13887/) (2025)
- [Deep Incomplete Multi-view Clustering with Distribution Dual-Consistency Recovery Guidance](https://mlsanity.com/paper/2503.11017/) (2025)

## Abstract (arXiv)

Incomplete multi-view clustering (IMVC) is typically evaluated by retraining separate models under different missing-view configurations. Evaluations indexed only by nominal missing rate can overlook differences in observation structure across missing-view protocols. We show that missing-data protocols with identical nominal missing rates can induce substantially different learning regimes, differing by approximately 50-fold in the proportion of fully observed samples. We formalize this phenomenon as protocol divergence, which quantifies structural disparities among missing-view protocols beyond marginal missing rates. Furthermore, we analyze support-gated reconstruction mechanisms and show that their optimization contribution is inherently limited by the frequency of eligible observations under explicit normalization and optimization conditions. Based on these observations, we propose CRAFT (Co-occurrence-free Robust Attention-masked Fusion Transformer), a train-once framework that combines representation learning with an architecture designed to process missing-view inputs. CRAFT combines (i) per-sample forward computation using each sample's observed views and shared parameters, and (ii) mask-aware fusion over nonempty observed-view subsets. The deployment evaluation starts from training data with all views available and reuses one final checkpoint per dataset and seed across missing-view protocols without retraining. Experiments on CUB and MultiFashion show that CRAFT achieves the strongest performance in 12 out of 13 information-matched settings. Additional deployment experiments across seven benchmarks and sixteen missing configurations demonstrate substantial computational savings through checkpoint reuse while maintaining competitive clustering performance. Code and evaluation tools: https://github.com/dk23lhl/CRAFT and https://github.com/dk23lhl/imvc-audit.