Tensor Decision Trees For High-Resolution Imaging Data

We propose a scalable, divide-and-conquer decision tree framework for high-resolution imaging regression, which operates in two stages to handle cases where the input image has high resolution. Our framework offers an efficient approach to managing the complexity of high-resolution tensor data while maintaining interpretability. We show that our model outperforms other tensor regression models using simulated data and provide results from the application to a neuroimaging dataset.