- Rating
- 3(+1)
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper proposes Federated Incomplete Multi-view Clustering via Prediction and Alignment (FIMCPA), addressing client-level view heterogeneity and sample-level view missingness. The framework combines masked cross-view predictive pretraining, sample-balanced contrastive alignment with shared-view relation verification, and observation-weighted model aggregation. A federated Lloyd refinement procedure is further employed for global clustering. Experiments on eight benchmark configurations demonstrate competitive clustering performance under varying missing rates and client compositions.
Critical strengths
-
The masked cross-view predictive pretraining strategy learns shared information from available observations without reconstructing genuinely missing views, providing a reasonable initialization for subsequent representation learning.
-
The framework jointly considers client-level view heterogeneity and sample-level missingness. The relation-verification mechanism further addresses potential false negatives in contrastive learning using shared-view neighborhood information.
-
The paper provides detailed theoretical and implementation analyses. Experiments across eight benchmarks, together with ablation studies, missing-rate evaluations, and client-composition analyses, demonstrate competitive clustering performance.
Critical weaknesses
-
The relation verifier determines neighborhood size based on an approximately balanced-cluster assumption in Eq. (5), while the experiments use class-stratified client partitions. Its effectiveness under label-skewed, naturally non-IID client distributions and imbalanced clusters remains unverified.
-
All evaluated client configurations retain full-view clients. The proposed framework is not tested when only partially overlapping view subsets are available across clients, leaving its effectiveness in the claimed arbitrary-view-subset setting insufficiently established.
-
FIMCPA employs global federated Lloyd refinement, whereas FMCSC and AFMCC retain different client-wise K-means procedures. Without evaluating all learned representations under the same clustering protocol, it is unclear how much of the improvement comes from representation learning rather than clustering refinement.
-
The ablation study in Table 2 does not independently evaluate sample-balanced contrastive alignment or observation-weighted aggregation, although both are central components of FIMCPA. Their individual contributions to the reported improvements remain unclear.
Questions for authors
-
Can the authors evaluate FIMCPA under label-skewed or naturally non-IID client partitions with imbalanced cluster distributions?
-
How does FIMCPA perform without full-view clients when the cross-client view co-observation graph remains connected?
-
Do the reported improvements persist when all methods use the same global clustering initialization and Lloyd refinement procedure?
-
Can the authors provide separate ablations comparing sample-balanced alignment with standard contrastive weighting and observation-weighted aggregation with conventional FedAvg?
8 comments
Response to Reviewer Ui4j W2/Q2: View Availability
W2/Q2: Learning without full-view clients. The original configurations retain full-view clients. To evaluate learning from partial overlap alone, we remove all full-view clients in two additional settings while preserving a connected view co-observation graph.
We evaluate P and P+S on all eight datasets using five paired random seeds for each setting, yielding 80 additional experiments. Ref is the default mixture (full:partial:single=2:1:1); P uses only partial-view clients (0:1:0); P+S mixes partial- and single-view clients equally (0:1:1). Partial-view clients acquire 2 to V−1 views. We retain class-stratified partitions, training budgets and an incomplete-sample rate of r=0.5 on multi-view clients. In every run, the view co-observation graph is connected: views are nodes, and an edge means that at least one sample observes both views on a client. All views are linked by such overlap paths, even though no client observes them all. ACC, NMI and ARI (%) are listed in rows for each dataset, reported as mean ± sample SD over five runs with different random seeds.
| Dataset | Metric | Ref | P | P+S |
|---|---|---|---|---|
| UCI | ACC | 85.92 ± 2.21 | 84.70 ± 1.12 | 79.74 ± 2.17 |
| NMI | 76.58 ± 1.98 | 74.76 ± 1.31 | 68.21 ± 2.10 | |
| ARI | 72.96 ± 3.32 | 70.77 ± 1.75 | 63.08 ± 3.28 | |
| HW | ACC | 90.20 ± 2.12 | 90.21 ± 0.80 | 85.62 ± 2.74 |
| NMI | 81.72 ± 2.71 | 81.41 ± 1.19 | 75.38 ± 1.96 | |
| ARI | 80.19 ± 3.52 | 79.93 ± 1.45 | 72.60 ± 2.92 | |
| Fashion | ACC | 89.81 ± 0.49 | 83.63 ± 1.77 | 78.41 ± 1.20 |
| NMI | 82.33 ± 0.65 | 76.35 ± 0.78 | 69.66 ± 0.69 | |
| ARI | 79.78 ± 0.89 | 71.29 ± 1.44 | 62.36 ± 1.18 | |
| Caltech-5V | ACC | 79.51 ± 3.21 | 79.06 ± 4.62 | 74.84 ± 2.21 |
| NMI | 66.94 ± 2.80 | 68.76 ± 1.80 | 58.84 ± 2.42 | |
| ARI | 63.65 ± 3.68 | 64.54 ± 2.67 | 55.51 ± 2.99 | |
| OutdoorScene | ACC | 73.90 ± 1.79 | 70.73 ± 0.87 | 65.77 ± 1.20 |
| NMI | 57.42 ± 1.78 | 54.26 ± 0.97 | 47.58 ± 1.41 | |
| ARI | 50.85 ± 2.45 | 46.75 ± 1.26 | 39.22 ± 1.80 | |
| CIFAR10-MV | ACC | 66.70 ± 2.62 | 66.57 ± 2.61 | 65.96 ± 1.24 |
| NMI | 54.61 ± 0.90 | 53.67 ± 1.46 | 52.50 ± 0.66 | |
| ARI | 45.01 ± 1.40 | 44.37 ± 2.53 | 43.42 ± 1.04 | |
| CCV | ACC | 24.45 ± 0.75 | 22.58 ± 0.99 | 19.95 ± 0.76 |
| NMI | 18.78 ± 0.18 | 17.34 ± 0.65 | 14.03 ± 0.43 | |
| ARI | 8.77 ± 0.30 | 8.07 ± 0.31 | 6.12 ± 0.33 | |
| MNIST-3V | ACC | 99.00 ± 0.06 | 98.57 ± 0.02 | 98.23 ± 0.06 |
| NMI | 96.93 ± 0.14 | 95.80 ± 0.05 | 94.97 ± 0.12 | |
| ARI | 97.78 ± 0.13 | 96.84 ± 0.05 | 96.11 ± 0.12 |
Without any full-view clients, P retains 97.32% of Ref ACC on average, showing that partial-view clients alone support effective clustering under the tested connected view-overlap patterns. P+S additionally includes single-view clients. Its results show that FIMCPA can still learn collaboratively from partial observations on the remaining clients when some clients cannot supply local cross-view prediction targets.
This capability stems from how MCPR exploits local co-observations. Each prediction task uses an actually observed view as its target and other observed views of the same sample as input; neither the sample nor its client needs all views. Removing full-view clients therefore does not remove cross-view predictive supervision. For example, clients with view subsets {1,2} and {2,3} construct their own local prediction tasks. Aggregation of updates to the shared view-2 encoder and fusion parameters connects these local learning tasks, allowing cross-view learning to exploit complementary observations distributed across clients.
Single-view clients do not construct cross-view prediction targets, but update the model through reconstruction and same-sample alignment, and receive and use parameters jointly learned by multi-view clients. Clients in P+S thus contribute to different objectives according to their available observations and collaborate through parameter sharing. Together, the P and P+S results show that full-view clients are not a prerequisite under the tested connected view-overlap patterns: distributed local co-observations still provide effective learning signals for FIMCPA.
Response to Reviewer Ui4j W3/Q3: Common Global Clustering
W3/Q3: Common global clustering. To assess representation quality under a matched final clustering procedure, we give FMCSC and AFMCC the same global initialization rule and Lloyd refinement as FIMCPA, and extend this evaluation to the remaining five baselines. The table reports ACC (%) as mean ± sample SD over five random seeds.
We retain training, features, default data partitions and masks. Local K-means provides centers and cluster sizes; count-weighted server-side K-means initializes global centers in each representation space (K-means++: 10 local/100 server restarts). All methods use at most 20 federated Lloyd updates, a convergence tolerance of , the same distance metric, tie-breaking rule and empty-cluster handling, and a single global Hungarian matching. Centers are updated using feature sums and sample counts aggregated across clients for each cluster.
FIMCPA uses ; FMCSC/AFMCC use learned representations with local clustering replaced. EFDMVC/FreeCSL/HSACC retain their K-means inputs (FreeCSL: observed-view consensus ). DSIMVC uses probabilities averaged over observed views; H2-FedMVC uses final MoE logits. The latter two serve as supplementary diagnostics of prediction-head outputs; H2-FedMVC retains label-assisted training.
| Dataset | FIMCPA | FMCSC | AFMCC | EFDMVC | FreeCSL | HSACC | DSIMVC | H2-FedMVC |
|---|---|---|---|---|---|---|---|---|
| UCI | 85.92 ± 2.21 | 73.07 ± 1.82 | 77.58 ± 0.92 | 76.09 ± 2.60 | 75.89 ± 3.96 | 68.51 ± 5.41 | 53.50 ± 3.36 | 77.86 ± 4.62 |
| HW | 90.20 ± 2.12 | 77.15 ± 1.92 | 78.34 ± 0.78 | 80.98 ± 4.14 | 82.77 ± 1.87 | 63.59 ± 5.78 | 50.40 ± 3.76 | 80.32 ± 2.54 |
| Fashion | 89.81 ± 0.49 | 84.80 ± 0.65 | 89.06 ± 0.17 | 84.26 ± 2.04 | 86.95 ± 4.68 | 81.55 ± 5.37 | 64.57 ± 1.12 | 85.97 ± 1.56 |
| Caltech-5V | 79.51 ± 3.21 | 71.99 ± 1.79 | 74.00 ± 4.30 | 65.64 ± 1.58 | 78.83 ± 3.76 | 33.09 ± 11.46 | 41.59 ± 3.42 | 72.09 ± 6.15 |
| OutdoorScene | 73.90 ± 1.79 | 62.40 ± 3.34 | 68.90 ± 2.18 | 61.38 ± 2.09 | 53.56 ± 4.01 | 45.42 ± 4.28 | 47.00 ± 2.90 | 57.48 ± 3.01 |
| CIFAR10-MV | 66.70 ± 2.62 | 53.03 ± 2.45 | 55.02 ± 2.87 | 51.55 ± 3.11 | 62.05 ± 4.46 | 34.85 ± 4.26 | 50.30 ± 1.00 | 41.18 ± 3.93 |
| CCV | 24.45 ± 0.75 | 19.50 ± 1.03 | 21.57 ± 0.47 | 20.19 ± 0.58 | 21.39 ± 0.90 | 14.95 ± 0.45 | 15.83 ± 0.21 | 15.92 ± 0.77 |
| MNIST-3V | 99.00 ± 0.06 | 94.77 ± 1.00 | 98.48 ± 0.02 | 98.27 ± 0.07 | 98.53 ± 0.09 | 72.06 ± 2.65 | 96.43 ± 3.93 | 95.17 ± 2.77 |
FMCSC and AFMCC now receive the same global coordination, yet FIMCPA exceeds them on all eight datasets, by 9.10 and 5.82 percentage points in ACC on average. FIMCPA has the highest mean ACC on all eight datasets. The mean margin over the strongest alternative per dataset is 3.74 points.
Under this protocol, FMCSC, AFMCC and all other baselines update centers using cluster statistics pooled across clients, receiving the same final global coordination. FIMCPA remains ahead, supporting the clustering quality of its learned common representations. Lloyd refinement changes centers and assignments on fixed representations without updating encoders, so this comparison controls the final clustering procedure.
During representation learning, MCPR removes the target view from both input paths and predicts it from other observed views, preventing direct copying of the target. Subsequent alignment connects same-sample view-specific and common representations, while relation verification uses shared-view neighborhood agreement to remove potential false negatives from the contrastive objective. These constraints shape the representation space before final clustering.
Global refinement updates common centers using cluster statistics aggregated across all clients, allowing each update to use information across clients rather than only samples from a single client. The complementary control uses the same FIMCPA representations and initial centers: client-wise refinement retains similar performance, while global refinement provides a further average improvement. For additional evidence on the effect of global refinement, please see our response to Reviewer PBoh, “Fixed-Representation Control.” Together, the comparisons show that FIMCPA's advantage cannot be explained solely by final global coordination; global refinement adds an average benefit on top of the learned common representations and shared initialization.
Response to Reviewer Ui4j W1/Q1: Non-IID Client Distributions
Thank you for your careful review and constructive comments. We report experiments on non-IID client distributions and learning without full-view clients, comparisons under matched global clustering, and independent weighting ablations. Below, we explain how the findings support FIMCPA's ability to learn from heterogeneous observations and the quality of its learned common representations.
W1/Q1: Non-IID client distributions. To evaluate learning under local class imbalance, we test Dirichlet non-IID partitions with α=1 on all eight datasets. We retain the original neighborhood and verification rules, together with the existing view heterogeneity and missing observations.
We report IID and non-IID ACC (%) as mean ± sample SD over five runs with different random seeds. ACC change denotes non-IID mean ACC minus IID mean ACC, in percentage points.
Dirichlet non-IID partitioning adds class-distribution heterogeneity to the existing view heterogeneity and missing observations. At the same missing-sample rate, locally underrepresented classes have fewer co-observed samples for prediction and fewer same-class pairs eligible for verification. The method must therefore learn from incomplete views with unequal numbers of observed samples across classes.
Caltech-5V shows the largest decrease (6.05 percentage points). Its 1,400 samples are distributed across clients with heterogeneous subsets of five views. Non-IID class proportions and missing observations further restrict the local samples available for prediction and relation verification, making learning more demanding with this limited sample pool. FIMCPA nevertheless achieves 73.46% ACC in this setting.
Equation (5) controls the candidate-neighborhood size; it does not assign samples to the same class. Under non-IID partitions, candidates must still satisfy mutual-neighbor agreement in every shared observed view, with at least two such views, before triggering negative removal. Even a verified relation is not added as a cross-sample positive. Positive pairs remain representations of the same sample, so their correspondence does not depend on local class proportions.
Candidate generation, negative removal and positive-pair construction therefore have distinct roles: neighborhood size defines candidates, cross-view verification determines negative removal, and sample identity defines positives. A change in neighborhood composition does not directly create cross-sample attraction. MCPR first learns from reconstruction and observed prediction targets; the subsequent contrastive stage combines reconstruction with relation-verified alignment.
Non-IID client distributions further increase the learning difficulty under view heterogeneity and missing observations. Under the Dirichlet setting with α=1, FIMCPA retains the original neighborhood and verification rules and preserves most of its IID performance, with an average ACC decrease of 3.81 percentage points across the eight datasets. Given the additional challenges posed by these combined sources of heterogeneity, we consider this average decrease acceptable in the evaluated setting. These results support FIMCPA's ability to adapt to the non-IID conditions tested here.