DeliciousMIL is a dataset designed for multi-label multi-instance learning (MIL) research. The dataset provides instance-level labels within its bag structure, a feature for advanced MIL algorithm development. It originates from the UCI Machine Learning Repository.
Use Cases
- Train models to predict bag-level labels from aggregated instance-level features and labels.
- Benchmark algorithms for multi-instance learning where each data 'bag' contains multiple instances.
- Develop methods for leveraging instance labels to improve the accuracy of bag-level multi-label predictions.
- Research the relationship between instance features and their corresponding labels within the MIL framework.
Strengths
- Designed for a specific and advanced machine learning paradigm (Multi-Instance Learning).
- Includes instance-level labels, which are not always available in standard MIL datasets.
Limitations
- Specific row count, column details, and data scale are unknown.
- Potential class distribution or feature imbalance is not documented.
Provenance
- Source
- UCI Machine Learning Repository
- Collection Method
- null
- Time Range
- null
- Freshness
- null
- Geography
- null