ﻻ يوجد ملخص باللغة العربية
Multi-view subspace learning (MSL) aims to find a low-dimensional subspace of the data obtained from multiple views. Different from single view case, MSL should take both common and specific knowledge among different views into consideration. To enhance the robustness of model, the complexity, non-consistency and similarity of noise in multi-view data should be fully taken into consideration. Most current MSL methods only assume a simple Gaussian or Laplacian distribution for the noise while neglect the complex noise configurations in each view and noise correlations among different views of practical data. To this issue, this work initiates a MSL method by encoding the multi-view-shared and single-view-specific noise knowledge in data. Specifically, we model data noise in each view as a separated Mixture of Gaussians (MoG), which can fit a wider range of complex noise types than conventional Gaussian/Laplacian. Furthermore, we link all single-view-noise as a whole by regularizing them by a common MoG component, encoding the shared noise knowledge among them. Such regularization component can be formulated as a concise KL-divergence regularization term under a MAP framework, leading to good interpretation of our model and simple EM-based solving strategy to the problem. Experimental results substantiate the superiority of our method.
Deep multi-view clustering methods have achieved remarkable performance. However, all of them failed to consider the difficulty labels (uncertainty of ground-truth for training samples) over multi-view samples, which may result into a nonideal cluste
Graph-based subspace clustering methods have exhibited promising performance. However, they still suffer some of these drawbacks: encounter the expensive time overhead, fail in exploring the explicit clusters, and cannot generalize to unseen data poi
Multi-typed objects Multi-view Multi-instance Multi-label Learning (M4L) deals with interconnected multi-typed objects (or bags) that are made of diverse instances, represented with heterogeneous feature views and annotated with a set of non-exclusiv
Many applications of machine learning, such as human health research, involve processing private or sensitive information. Privacy concerns may impose significant hurdles to collaboration in scenarios where there are multiple sites holding data and t
The information bottleneck principle provides an information-theoretic method for representation learning, by training an encoder to retain all information which is relevant for predicting the label while minimizing the amount of other, excess inform