In many modern applications from, for example, bioinformatics and computer vision, samples have multiple feature representations coming from different data sources. Multiview learning algorithms try to exploit all these available information to obtain a better learner in such scenarios. In this paper, we propose a novel multiple kernel learning algorithm that extends kernel k -means clustering to the multiview setting, which combines kernels calculated on the views in a localized way to better capture sample-specific characteristics of the data. We demonstrate the better performance of our localized data fusion approach on a human colon and rectal cancer data set by clustering patients. Our method finds more relevant prognostic patient groups than global data fusion methods when we evaluate the results with respect to three commonly used clinical biomarkers.

First provide a summary of the paper, and then address the following criteria: Quality, clarity, originality and significance. This paper generalizes multi-kernel k-means clustering (solved via relaxation) to the case where each clustered item (here, a person) gets an item-specific set of weights over the multiple kernels, rather than the traditional, shared, global weighting of the kernels. Using TCGA (cancer) data, with 3 modalities, they demonstrate that this generalization yields better clusterings than the traditional (global approach), when measured against 3 bronze standard clusterings arising from known clinical clusters. The writing is clear, making for an easy read. Although this is a somewhat incremental-seeming tweak, I think it was clever, with the potential to actually be used (rather than lost in the NIPS archives), and therefore of some significance. Other comments: In the introduction you mention that k-means is susceptible to local minima, and then use this to motivate the relaxation approach.

A very common heuristics to solve this problem is the Lloyd's algorithm [

We propose a novel type of balanced clustering algorithm to approximate attention.

Neural Information Processing Systems http://nips.cc/

First provide a summary of the paper, and then address the following criteria: Quality, clarity, originality and significance. In this paper the authors propose a novel bi-clustering approach based on a message passing algorithm. The motivation is that most current bi-clustering techniques overcome the computational difficulty of the problem by performing greedy local optimisations. In this paper, the authors propose to overcome this by defining a global likelihood function (eq 1) and then maximising an approximation to this function via message passing. Quality: this is a high quality paper.

Table 1 reports the results of a controlled experiment evaluating different components in our framework for auxiliary label generation. Positive correspondences are generated by matching pixels across different augmentations of the same image. With respect to the clustering algorithm, K-means performs better than DBSCAN (#4 vs. #5), which is We show qualitative results, comparing different feature extractors in Figure 1. DBSCAN is limited by the memory and computational complexity. Corresponding qualitative results are shown in Figure 3. Tables 3-5 show We observe the best performance when 5% outliers are removed.

For this problem, we propose a novel generalization of random walk, diffusion, or smooth function methods in the literature to a convex p-norm cut function.