Approximation Algorithms for Cascading Prediction Models.

We present an approximation algorithm that takes a pool of pre-trained modelsas input and produces from it a cascaded model with similar accuracy but loweraverage-case cost. Applied to state-of-the-art ImageNet classification models,this yields up to a 2x reduction in floating point multiplications, and up to a6x reduction in average-case memory I/O. The auto-generated cascades exhibitintuitive properties, such as using lower-resolution input for easier imagesand requiring higher prediction confidence when using a computationally cheapermodel.