2019/09/11 by Stephen Gould, Richard Hartley, Dylan Campbell +1 · 1 voice · 52 citations
Computer Science · Materials Science · Physics and Astronomy · #Class (philosophy) #Declarative programming #Deep learning #Function (biology) #Machine Learning in Materials Science #Model Reduction and Neural Networks #Programming paradigm #Software #Stochastic Gradient Optimization Techniques #cs.AI #cs.CV #cs.LG
paper · pdf · doi:10.1109/tpami.2021.3059462
published in IEEE Transactions on Pattern Analysis and Machine Intelligence 44(8), 1 (IEEE Computer Society)
arxiv published 2019/09/11 · openalex created_date 2019/09/19 · arxiv created 2020/02/27 · openalex publication_date 2021/01/01 · arxiv updated 2021/08/20 · openalex updated_date 2026/08/05
We explore a class of end-to-end learnable models wherein data processing nodes (or network layers) are defined in terms of desired behavior rather than an explicit forward function. Specifically, the forward function is implicitly defined as the solution to a mathematical optimization problem. Consistent with nomenclature in the programming languages community, we name these models deep declarative networks. Importantly, it can be shown that the class of deep declarative networks subsumes current deep learning models. Moreover, invoking the implicit function theorem, we show how gradients can be back-propagated through many declaratively defined data processing nodes thereby enabling end-to-end learning. We discuss how these declarative processing nodes can be implemented in the popular PyTorch deep learning software library allowing declarative and imperative nodes to co-exist within the same network. We also provide numerous insights and illustrative examples of declarative nodes and demonstrate their application for image and point cloud classification tasks.