Sparse Compositional Metric Learning
This addresses metric learning for classification tasks, offering a flexible framework that is incremental in improving efficiency and generalization.
The paper tackles metric learning by learning a sparse combination of locally discriminative metrics, which reduces parameters and generalizes to new data. Empirical results show superior classification performance and scalability compared to state-of-the-art methods.
We propose a new approach for metric learning by framing it as learning a sparse combination of locally discriminative metrics that are inexpensive to generate from the training data. This flexible framework allows us to naturally derive formulations for global, multi-task and local metric learning. The resulting algorithms have several advantages over existing methods in the literature: a much smaller number of parameters to be estimated and a principled way to generalize learned metrics to new testing data points. To analyze the approach theoretically, we derive a generalization bound that justifies the sparse combination. Empirically, we evaluate our algorithms on several datasets against state-of-the-art metric learning methods. The results are consistent with our theoretical findings and demonstrate the superiority of our approach in terms of classification performance and scalability.