Neuro-symbolic Neurodegenerative Disease Modeling as Probabilistic Programmed Deep Kernels
This work addresses the problem of early and accurate neurodegenerative disease prediction for medical AI applications, offering improvements in interpretability and data-efficiency, though it appears incremental as it builds on existing Bayesian and kernel methods.
The authors tackled personalized prediction of neurodegenerative diseases by developing a probabilistic programmed deep kernel learning approach, which outperformed deep learning in accuracy and timeliness for Alzheimer's disease prediction without requiring clinical labels for training.
We present a probabilistic programmed deep kernel learning approach to personalized, predictive modeling of neurodegenerative diseases. Our analysis considers a spectrum of neural and symbolic machine learning approaches, which we assess for predictive performance and important medical AI properties such as interpretability, uncertainty reasoning, data-efficiency, and leveraging domain knowledge. Our Bayesian approach combines the flexibility of Gaussian processes with the structural power of neural networks to model biomarker progressions, without needing clinical labels for training. We run evaluations on the problem of Alzheimer's disease prediction, yielding results that surpass deep learning in both accuracy and timeliness of predicting neurodegeneration, and with the practical advantages of Bayesian nonparametrics and probabilistic programming.