Demystifying Differentiable Programming: Shift/Reset the Penultimate Backpropagator
This work addresses the need for more flexible and efficient training methods in machine learning as architectures become more complex, offering a foundational improvement for developers and researchers.
The paper tackles the challenge of generalizing differentiable programming for arbitrary parameterized computations by revealing a connection between reverse-mode automatic differentiation and delimited continuations, enabling an efficient implementation that matches the performance of frameworks like TensorFlow while maintaining the expressiveness of PyTorch.
Deep learning has seen tremendous success over the past decade in computer vision, machine translation, and gameplay. This success rests in crucial ways on gradient-descent optimization and the ability to learn parameters of a neural network by backpropagating observed errors. However, neural network architectures are growing increasingly sophisticated and diverse, which motivates an emerging quest for even more general forms of differentiable programming, where arbitrary parameterized computations can be trained by gradient descent. In this paper, we take a fresh look at automatic differentiation (AD) techniques, and especially aim to demystify the reverse-mode form of AD that generalizes backpropagation in neural networks. We uncover a tight connection between reverse-mode AD and delimited continuations, which permits implementing reverse-mode AD purely via operator overloading and without any auxiliary data structures. We further show how this formulation of AD can be fruitfully combined with multi-stage programming (staging), leading to a highly efficient implementation that combines the performance benefits of deep learning frameworks based on explicit reified computation graphs (e.g., TensorFlow) with the expressiveness of pure library approaches (e.g., PyTorch).