Multi-Instance Multi-Label Learning

Abstract

In this paper, we propose the MIML (Multi-Instance Multi-Label learning) framework where an example is described by multiple instances and associated with multiple class labels. Compared to traditional learning frameworks, the MIML framework is more convenient and natural for representing complicated objects which have multiple semantic meanings. To learn from MIML examples, we propose the MimlBoost and MimlSvm algorithms based on a simple degeneration strategy, and experiments show that solving problems involving complicated objects with multiple semantic meanings in the MIML framework can lead to good performance. Considering that the degeneration process may lose information, we propose the D-MimlSvm algorithm which tackles MIML problems directly in a regularization framework. Moreover, we show that even when we do not have access to the real objects and thus cannot capture more information from real objects by using the MIML representation, MIML is still useful. We propose the InsDif and SubCod algorithms. InsDif works by transforming single-instances into the MIML representation for learning, while SubCod works by transforming single-label examples into the MIML representation for learning. Experiments show that in some tasks they are able to achieve better performance than learning the single-instances or single-label examples directly.

64 pages, 10 figures; Artificial Intelligence, 2011

Cited by in corpus (42)

• Multiple Instance Learning: A Survey of Problem Characteristics and Applications
• Learning to Discover Multi-Class Attentional Regions for Multi-Label Image Recognition
• An End-to-End Deep Learning Histochemical Scoring System for Breast Cancer Tissue Microarray
• On perfect clustering of high dimension, low sample size data
• Provably Consistent Partial-Label Learning
• On Classification with Bags, Groups and Sets
• A snapshot on nonstandard supervised learning problems: taxonomy, relationships and methods
• Weighting Scheme for a Pairwise Multi-label Classifier Based on the Fuzzy Confusion Matrix
• Learning Category Correlations for Multi-label Image Recognition with Graph Networks
• Weakly-supervised Dictionary Learning
• Multi-Instance Multi-Label Learning for Gene Mutation Prediction in Hepatocellular Carcinoma
• A Correction Method of a Binary Classifier Applied to Multi-label Pairwise Models
• Segregated Temporal Assembly Recurrent Networks for Weakly Supervised Multiple Action Detection
• Learning and Interpreting Multi-Multi-Instance Learning Networks
• An Attention Mechanism for Musical Instrument Recognition
• Dynamic classifier chains for multi-label learning
• Jointly Extracting Relations with Class Ties via Effective Deep Ranking
• Fast Multi-Instance Multi-Label Learning
• Multi-label Ranking: Mining Multi-label and Label Ranking Data
• Using Multiple Instance Learning to Build Multimodal Representations
• MIML library: a Modular and Flexible Library for Multi-instance Multi-label Learning
• Cross-Class Relevance Learning for Temporal Concept Localization
• Extreme Multi-label Classification from Aggregated Labels
• AMI-Net+: A Novel Multi-Instance Neural Network for Medical Diagnosis from Incomplete and Imbalanced Data
• Multi-View Multi-Instance Multi-Label Learning based on Collaborative Matrix Factorization
• Weakly Supervised Learning Meets Ride-Sharing User Experience Enhancement
• Bidirectional Loss Function for Label Enhancement and Distribution Learning
• Learning to Learn and Predict: A Meta-Learning Approach for Multi-Label Classification
• Generative-Discriminative Complementary Learning
• VICSOM: VIsual Clues from SOcial Media for psychological assessment
• Evaluation of Joint Multi-Instance Multi-Label Learning For Breast Cancer Diagnosis
• Weakly Supervised Person Re-Identification
• Novelty Detection Under Multi-Instance Multi-Label Framework
• Weakly Supervised Image Annotation and Segmentation with Objects and Attributes
• Convex and Scalable Weakly Labeled SVMs
• Weakly-Supervised Multi-Person Action Recognition in 360${}^{\circ }$ Videos
• MIML-FCN+: Multi-instance Multi-label Learning via Fully Convolutional Networks with Privileged Information
• Discovering Multi-Label Actor-Action Association in a Weakly Supervised Setting
• Reconstruction Regularized Deep Metric Learning for Multi-label Image Classification
• Multi-typed Objects Multi-view Multi-instance Multi-label Learning
• Deep Ranking Based Cost-sensitive Multi-label Learning for Distant Supervision Relation Extraction
• Inferring Restaurant Styles by Mining Crowd Sourced Photos from User-Review Websites