Input Adaptive Bayesian Model Averaging
It addresses the challenge of model combination in heterogeneous data for applications such as personalized medicine and fraud detection, representing an incremental improvement over existing adaptive methods.
The paper tackles the problem of combining multiple candidate models for prediction in heterogeneous settings by proposing Input Adaptive Bayesian Model Averaging (IA-BMA), which assigns model weights based on inputs and yields more accurate and better-calibrated predictions across tasks like cancer treatment and fraud detection.
This paper studies prediction with multiple candidate models, where the goal is to combine their outputs. This task is especially challenging in heterogeneous settings, where different models may be better suited to different inputs. We propose input adaptive Bayesian Model Averaging (IA-BMA), a Bayesian method that assigns model weights conditional on the input. IA-BMA employs an input adaptive prior, and yields a posterior distribution that adapts to each prediction, which we estimate with amortized variational inference. We derive formal guarantees for its performance, relative to any single predictor selected per input. We evaluate IABMA across regression and classification tasks, studying data from personalized cancer treatment, credit-card fraud detection, and UCI datasets. IA-BMA consistently delivers more accurate and better-calibrated predictions than both non-adaptive baselines and existing adaptive methods.