• Corpus ID: 49526066

A probabilistic constrained clustering for transfer learning and image category discovery

  title={A probabilistic constrained clustering for transfer learning and image category discovery},
  author={Yen-Chang Hsu and Zhaoyang Lv and Joel Schlosser and Phillip Odom and Zsolt Kira},
Neural network-based clustering has recently gained popularity, and in particular a constrained clustering formulation has been proposed to perform transfer learning and image category discovery using deep learning. The core idea is to formulate a clustering objective with pairwise constraints that can be used to train a deep clustering network; therefore the cluster assignments and their underlying feature representations are jointly optimized end-to-end. In this work, we provide a novel… 

Tables from this paper

Joint Representation Learning and Novel Category Discovery on Single- and Multi-modal Data
A generic, end-to-end framework to jointly learn a reliable representation and assign clusters to unlabelled data is presented and Winner-Take-All hashing algorithm is employed on the shared representation space to generate pairwise pseudo labels for unlabelling data to better predict cluster assignments.


Deep Image Category Discovery using a Transferred Similarity Function
This paper presents a novel end-to-end network to map unlabeled images to categories as a clustering network and proposes that this network can be learned with contrastive loss which is only based on weak binary pair-wise constraints.
Neural network-based clustering using pairwise constraints
A novel strategy to utilize the contrastive criteria for pushing data-forming clusters directly from raw data, in addition to learning a feature embedding suitable for such clustering, which beats the conventional two-stage method by a significant margin.
Learning to cluster in order to Transfer across domains and tasks
A novel method to perform transfer learning across domains and tasks, formulating it as a problem of learning to cluster with state of the art results for the challenging cross-task problem, applied on Omniglot and ImageNet.
Semi-Supervised Kernel Mean Shift Clustering
This work proposes a semi-supervised framework for kernel mean shift clustering (SKMS) that uses only pairwise constraints to guide the clustering procedure and shows the advantages of SKMS by evaluating its performance on various synthetic and real datasets while comparing with state-of-the-art semi- supervised clustering algorithms.
Integrating constraints and metric learning in semi-supervised clustering
Experimental results demonstrate that the unified approach produces better clusters than both individual approaches as well as previously proposed semi-supervised clustering algorithms.
Image Clustering Using Local Discriminant Models and Global Integration
This paper proposes a new image clustering algorithm, referred to as clustering using local discriminant models and global integration (LDMGI), and shows that LDMGI shares a similar objective function with the spectral clustering (SC) algorithms, e.g., normalized cut (NCut).
On constrained spectral clustering and its applications
This paper presents a more natural and principled formulation of constrained spectral clustering, which explicitly encodes the constraints as part of a constrained optimization problem, and demonstrates an innovative use of encoding large number of constraints: transfer learning via constraints.
Large-Scale Object Discovery and Detector Adaptation from Unlabeled Video
This work proposes a fully automatic approach for object mining from video which builds upon a generic object tracking approach and proposes a novel approach for unsupervised object discovery by appearance-based clustering.
Large Scale Spectral Clustering with Landmark-Based Representation
This paper proposes a novel approach, called Landmark-based Spectral Clustering (LSC), for large scale clustering problems, where p representative data points are selected as the landmarks and the spectral embedding of the data can be efficiently computed with the landmark-based representation.
Very Deep Convolutional Networks for Large-Scale Image Recognition
This work investigates the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting using an architecture with very small convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers.