PMODE: Theoretically Grounded and Modular Mixture Modeling
This work addresses mixture modeling for researchers and practitioners, offering a theoretically grounded and modular approach that is incremental in scaling existing methods.
The authors tackled the problem of mixture modeling by introducing PMODE, a modular framework that partitions data and fits separate estimators to each subset, achieving near-optimal rates and handling components from different distribution families. As an application, MV-PMODE scaled a theoretical approach to high-dimensional settings, performing competitively against deep baselines on CIFAR-10 anomaly detection.
We introduce PMODE (Partitioned Mixture Of Density Estimators), a general and modular framework for mixture modeling with both parametric and nonparametric components. PMODE builds mixtures by partitioning the data and fitting separate estimators to each subset. It attains near-optimal rates for this estimator class and remains valid even when the mixture components come from different distribution families. As an application, we develop MV-PMODE, which scales a previously theoretical approach to high-dimensional density estimation to settings with thousands of dimensions. Despite its simplicity, it performs competitively against deep baselines on CIFAR-10 anomaly detection.