Stochastic Variational Deep Kernel Learning
This work addresses scalability limitations in deep kernel learning for practitioners needing efficient, flexible models on large datasets, representing a hybrid advancement rather than a paradigm shift.
The authors tackled the challenge of scaling deep kernel learning to large datasets by proposing a novel model with stochastic variational inference that enables classification, multi-task learning, and additive covariance structures. They demonstrated improved performance over deep networks, SVMs, and scalable Gaussian processes on benchmarks like CIFAR and ImageNet, including handling 6 million training points on an airline delay dataset.
Deep kernel learning combines the non-parametric flexibility of kernel methods with the inductive biases of deep learning architectures. We propose a novel deep kernel learning model and stochastic variational inference procedure which generalizes deep kernel learning approaches to enable classification, multi-task learning, additive covariance structures, and stochastic gradient training. Specifically, we apply additive base kernels to subsets of output features from deep neural architectures, and jointly learn the parameters of the base kernels and deep network through a Gaussian process marginal likelihood objective. Within this framework, we derive an efficient form of stochastic variational inference which leverages local kernel interpolation, inducing points, and structure exploiting algebra. We show improved performance over stand alone deep networks, SVMs, and state of the art scalable Gaussian processes on several classification benchmarks, including an airline delay dataset containing 6 million training points, CIFAR, and ImageNet.